Research Article Accuracy-Energy Configurable Sensor Processor …downloads.hindawi.com/journals/tswj/2014/546563.pdf · · 2015-11-23Accuracy-Energy Configurable Sensor Processor

Research ArticleAccuracy-Energy Configurable Sensor Processor andIoT Device for Long-Term Activity Monitoring in Rare-EventSensing Applications

Daejin Park and Jeonghun Cho

School of Electronics Engineering Kyungpook National University Daegu 702-701 Republic of Korea

Correspondence should be addressed to Jeonghun Cho jchoeeknuackr

Received 10 September 2014 Accepted 25 October 2014 Published 14 December 2014

Academic Editor Hai Jiang

Copyright copy 2014 D Park and J Cho This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

A specially designed sensor processor used as a main processor in IoT (internet-of-thing) device for the rare-event sensingapplications is proposedThe IoT device including the proposed sensor processor performs the event-driven sensor data processingbased on an accuracy-energy configurable event-quantization in architectural level The received sensor signal is converted into asequence of atomic events which is extracted by the signal-to-atomic-event generator (AEG) Using an event signal processingunit (EPU) as an accelerator the extracted atomic events are analyzed to build the final event Instead of the sampled raw datatransmission via internet the proposed method delays the communication with a host system until a semantic pattern of the signalis identified as a final event The proposed processor is implemented on a single chip which is tightly coupled in bus connectionlevel with a microcontroller using a 018 120583m CMOS embedded-flash process For experimental results we evaluated the proposedsensor processor by using an IR- (infrared radio-) based signal reflection and sensor signal acquisition system We successfullydemonstrated that the expected power consumption is in the range of 20 to 50 compared to the result of the basement in caseof allowing 10 accuracy error

1 Introduction

In recent years rare-event sensing systems [1] have been usedin IoT-driven applications [2] These systems feature internetconnectivity low-cost very slow event-to-event durationand long-lasting requirements The IoT devices are used foractivity monitoring [3] human sense interface [4] securitymonitoring and medical applications [5] The requirementfor an extremely long lifetime is a critical issue in battery-operated IoT devices with a wireless connectivity

The general-purpose microcontroller (MCU) has beenwidely used as a main processor of IoT-driven systems forsensor signal acquisition and data processing which arenot especially suited for rare-event sensing applications Theinefficiency of conventionalMCUs in the sensing applicationshas been introduced in various literature [1 6 7] thatreveal the key requirements of sensing application-specificarchitecture and processing methods The latest studies ofprocessor design for sensing applications are primarily based

on architectural approaches that consider an event-drivennature [8] in extracting informative features from raw sen-sory data by observing over long periods of time

Figure 1 illustrates basic operation and power consump-tion in a main sensor processor of the IoT device A timelysampling operation to measure the transition of the targetenvironment begins with activation of the sensor An analog-to-digital converter (ADC) and level comparators performanalog data conversion which generates digitized sampledata and an interrupt request to execute the user-definedsubroutines for digital signal processing The CPU in theMCU is woken up to execute user-programmed interruptservice routines (ISRs) which are software code to analyze thequantized sensor data Finally the gathered data is transferredvia internet connection to host systems

The general-purpose MCU-based IoT device consumesinefficiently operating power in an iterativemannerDiscrete-time-based sampling and signal processing are executediteratively during the entire period of sensor signal transition

Hindawi Publishing Corporatione Scientific World JournalVolume 2014 Article ID 546563 16 pageshttpdxdoiorg1011552014546563

2 The Scientific World Journal

Power

Sensor

Sampling (ADC CMP)

CPU execution

Voltage

Long-term

Long-term

no activity

Triggered event

Internet connection

Recognized

activity

PsensorPsampling

Pup

Tduty

Tsensor Tsampling

Tup

TwakeupTCPU

Tinternet

2mA

16mA

600120583A

350120583A

t

t

Curr

entlowast

5V

Sensor signal s(t)

Atomic event (aevi)

event (evk)

aev0 aevi

ev0

ev1

Pleakage

Figure 1 Power consumption for sensor signal sampling andprocessing in long-term activity monitoring

even during the period of long-term sleep Although thisweakness in terms of the power consumption can be resolvedusing several programming techniques where the interrupthandlers consider the sensor signal behavior this is notformal approach that covers the general cases of signalcharacteristics

Several approaches [6 9ndash11] regarding the behavior ofsensing applications have been introduced in sensor-basedapplication for low power operations especially in activitysensor sensor processor andwireless sensor nodes Howeverthe use of the discrete-time-based operation is restrictedbecause the most feasible devices rely on commercial off theshelf (COTS) MCUs attached with several sensors Thesedevices attempt to control the ratio of the sleep mode tothe active mode by switching the operation mode to lowerthe power consumption according to being with or withoutthe sensory data which are continuously determined by thediscrete-time-based monitoring of the signal

2 Motivation

To address this limitation of the conventional digital sys-tem architecture by using the discrete-time-based sen-sor data processing method we propose an event-drivensystem architecture that modifies traditional digital systemdesign We present a theoretical framework to implement anevent-driven sensor processor for general rare-event sensingapplications by analyzing the system operations

21 Event-Space Signal Representation Our main researchbegins with an event-space representation of the sig-nal instead of the digital data space domain Figure 2(a)

illustrates fundamental signal elements using attributesof interest and Figure 2(b) shows the event data spacetransition-based representation which describes in detail therelationship between the featured points of the sensed signal

The extracted features of the sensed signal are encodedinto the elapsed time between events and informative valuesuch as voltage level and edge phase crossing the trigger pointof the signalThe fundamental event defined which is definedas an atomic event with the most important informationprovides a signal representation on an abstract level andreduces the computational complexity in performing basicdata processing for extracted informative features of interestThe collected atomic events include partial information inthe original signal that specifies whether the desired featuredpoints of the signal are present

22 Event-Quantization with Accuracy Error The event-quantization concept extends the time-quantization methodfor signal representation that uses elapsed time to enhance theconventional data sampling and processing method Time-quantization monitors only the specific conditions of thesignal transition and captures the time-stamps The event-quantization method also determines whether the specifiedcharacteristics of the signal exist Figures 2(c) and 2(d)illustrate the difference between the time-sampling andevent-quantization methods with an accuracy error whichmonitors only the presence of the featured points of interest

In terms of accuracy there are two types of operationstiming accuracy and data resolution accuracy [12] Theformer is dependent on the sampling frequency comparedto the received signal bandwidth and the latter is derivedfrom the representative resolution of the sampled data Thetiming resolution is dependent on the clock-duty resolutionto resolve the timing window of the processing operationsaccurately

Higher timing resolution for the duration measurementrequires the accurate data processing in the time domainHowever inmost cases the data value resolution of the targetsystem is tightly required but timemdashrelated specificationincluding the response time is relatively allowable in a certainrange of error In particular in rare-event sensing applica-tions we assume that the timing measurement accuracy forthe elapsed time between the arrived signal events can beperformed with inaccurate operating clock relatively

Although the events contain inaccurate time informationthe final event receiver (for example a human interface) isunable to identify the error differences compared to the idealdata within a certain error range [13]

23 Accuracy Configurable Signal-to-Event Conversion Theevent-based approach with a certain amount of accuracyerror is described by the proposed event-driven sensor dataprocessing flow as shown in Figure 2(e) The input signalis monitored with specified interest-of-signal characteristics119882(119896) to generate the specific atomic events aev

119894of the signal

The set of atomic events during the specified region of thesignal are traced into the tracer memory as an event vector997888997888997888rarrAEV which contains the sequence of the atomic events and

The Scientific World Journal 3

Attribute and time of interestrepresentation

(1) thk attribute of interest(2) Φ edge phase(3) eti elapsed time

Distance

s(t)

s(t)

thc

thb

tha

taev0aev1 aev2 aev3 aev4 aev5

= aevi | aevi = ⟨(thk Φ) eti⟩minusminusminusrarrAEV

(a) Attribute and its corresponding elapsed time representation

t

Sens

or si

gnal

sp

ace

t

Ea

Eb

Ec

EdEven

t dat

a spa

ce

= ⟨Ec et4⟩aev4

= ⟨Ed et5⟩aev5

et0 et1 et2 et3 et4 et5

= ⟨Ea et0⟩aev0= ⟨Eb et1⟩aev1


(b) An example of event-space representation for incoming sensor signal

High accurate clock

Long-term no activity

Elapsed time-based representation

s(t)

OSC

tdi

(thk eti)

(c) Accurate time sampling

Low accurate clocks(t)

OSC

aevi

aeva aevb aevinfin aevc aevc

(d) Event sampling with accuracy error

gttimes

times

Δe

z(t)

z(k)

Event tracing

Interest attribute rule

(threshold levels)

Atomic event

EventidentificationAtomic

event vector

Expected rule of atomic events

Received vectorAccuracy

rangeAccuracy-controlled event recognition

Cs

Ds

EVj

Z(k)|k=⟨sel⟩

= (eyi et i)

= (Ei eti)

Δe

minusminusminusrarrAEV


minusminusrarrRV

(e) Overall signal processing model with accuracy-controlled signal-to-event conversion

Figure 2 Event-space quantization and accuracy-configurable event-driven sensor processing flow

the time distance relationship between the atomic eventsThetraced event vector identifies the approximate result119885(119896) as afinal event by comparing it to the expected rules of the atomicevents

These approximation approaches enable us to reduce thecomputational complexity in order to manipulate a largeamount of collected sensing data As a result power con-sumption will be reduced For applications related to human

interaction an approximation approach enables developersto design the computational block using smaller hardwareresources while providing sufficient performance in limitedresolution of the accuracy

In accuracy-controlling approaches defined from thespecifications our study focused on the data-representationresolution the timing resolution of the sampling frequencyand the response time as a delay time [14] This enables


Time

Data quantizationData

Dk = Δs middot k

ti = ts lowast i

s(t)

∙ Polling operation (active)∙ Capturing change of status

Event-quantization

OSC

Time

Data

s(t)

Tk = Δt middot k

li = ls lowast i

⟨li Tk⟩

∙ Capturing signal shape∙ Determining presence of features

Time

Data

Time-quantization

OSC

s(t)

Tk = Δt middot k

li = ls lowast i

∙ Interrupt operation (passive)∙ Capturing time-stamp

(a) Target

Accurate datatime-quantization

Data

Dk = Δs middot k

s(t)

TimeTk = Δt middot k

∙ Higher sampling frequency anddata frequency

Time

Data

Inaccurate datatime-

quantization

Feature representation as a range of value

s(t)

Δs middot m lt Dk lt Δs middot j + Δe

Δt middot k lt Tk lt Δt middot j + Δe

∙ Instead of data itself determining thepresence of datatime in specific range

Time

Data

Inaccurate data quantization

OSCData approximation approach

s(t)

Dk = Δs middot k + Δe

∙ Decreasing the number ofdata bit representation

(b) Accuracy

t

t

t

OSC

t

tEvent-

capturedEvent-

triggered

Maximum accuracy error

s(t)tha

aeidealae0

dclk

Δosc Δev

eclk

ΔaΔe

max(Δosc) = dclk + eclkmax(Δe) = Δa + Δosc + Δev

ae0998400

(c) Accuracy error

Figure 3 Category according to sampled target and its accuracy

the configuration of the operation accuracy in the processorarchitecture level according to the abstraction level of theproposed event-quantization approach

3 Related Work and Constraints

To overcome the weakness of the frequent CPU wake-up in the discrete-time sampling continuous time signalprocessing techniques [15 16] have been proposed If a certaincondition of the signal status such as the voltage level ata specific time is matched with the user-defined condition[17 18] the time value at the triggered condition is sampledand quantized [19] by the selective method which also helpsto reduce operational power [20]

The continuous time sampling method illustrated in thesecond graph of Figure 3(a) requires additional hardware

resources including a dedicated oscillator and high-accuratetimer block tomeasure the elapsed time value instead of datasampling in high resolution Event-driven signal processingbased on the time-quantization method requires hardwareoverhead andmore computational time for the time-distancecalculation which gives rise to additional power consump-tion The required power and hardware resource overheadwhich are needed to compensate for reduced wake-up powerconsumptionmust be considered in order to achieve benefitsin total energy efficiency due to hardware-energy trade-off

The hybrid method which uses a level triggered systemwake-up and continuous time sampling scheme monitorsimportant signal transitions and performs detailed analysisusing the discrete-time sampling method which involvesdigital signal processing for second detail signal analysisTheproposed event-driven sensor data processing method triesto capture the signal shape instead of the elapsed time-stamp


Rare-event applications dmk ≫ em998400 m

em998400 m

thi = Δ lowast i

thi = Δ lowast i

thi = Δ lowast i

n gt m n asymp m

n ≫ m rare-event applicationsLi = Δ lowast i

tk = Δt lowast k dmktm = Δt lowast m tm998400 = Δt lowast m998400

s(t)

s(t)

s(t)

Time

Time

TimeLong-term no activity


Time

middot middot middot

Figure 4 Wake-up frequency for data sampling and event-drivensampling

at the triggered point as illustrated in the third graph ofFigure 3(a)

The trade-offs in terms of energy and accuracy havebeen studied widely [21 22] To obtain long lifetime oper-ations under limited battery power [23] the latest researchintroduces inaccurate computation techniques [12 24] withapproximation-based hardware designs as described in thesecond graph of Figure 3(b)

The proposed sensing processor for the rare-event sens-ing applications adopts the event-driven approach of thecontinuous time-based sampling method Inaccurate time-datamanipulation as shown in the third graph of Figure 3(b)reduces computational complexity and sampling resolutionby determining the presence of featured events in the specificrange The level that allows an accuracy error in the time-stamp measurement is depicted in Figure 3(c) The level canbe adjusted by making the trade-off between the processingenergy consumption and the operating specification

Figure 4 shows the difference between the discrete-timesamples and the featured events of interest with the commonshape of the rare-event sensor signal Event sources such ashand gesture proximity and object activity generate signalpulses for which the distance between featured points of thesignal is very long The number of data samples (119899) is greaterthan the number of events (119898) In this work we assumed anapplication-specific constraint of rare-event characteristicswhich result in a small number of events compared to thenumber of data samples

The event-quantization accuracy depending on the res-olution of the elapsed time-stamp is described as 119890

1198981015840119898

inFigure 4 The rare-event sensing applications for which theevent-to-event duration is relatively larger than the accuracyerror have the following applications-specific constraints

119889119898119896

≫ 1198901198981015840119898 (1)

With these application-specific constraints the eventidentification accuracy error caused by the inaccurate time-stamp measurement clock is relatively insensitive derived

from (1) The recognized event observer such as humaneye allows a certain amount of inaccuracy in identifying ameaning of the events which are constructed by the proposedinaccurate event-driven sensor processor

The proposed sensor processor is designed with theseapplication-specific constraints by reducing the accuracyof the time-stamp measurement clock decreasing the bitwidth of the timer block to capture the time-stamps anddecreasing the operational complexity of the time-to-timedistance measurement blocks which are specially imple-mented as a dedicated accelerator for event recognition in theimplemented hardware

4 Proposed Architecture

41 Data-Time Sampling with Accuracy Error The first stageof the sensor processor is a sampler that gathers the time-variant information from the received signal The conven-tional sampling method in Figure 5(a) attempts to collect allinformation in discrete-time from the target signal There isno need to hold the time-stamp dataThe uniformly sampledset in the timedomain is described in the following definition

Definition 1 Given continuous signal 119904(119905) let 119905119904be a fixed

sampling time and 119878unitime = 1199041 1199042 119904

119899 a uniformly

sampled set in time domain 119879 = 1199051 1199052 119905

119899 A sampled

data 119904119894isin 119878 and its data quantization result with error Δ

119889by

data quantization function DQ are defined as follows

119889119894plusmn Δ119889= DQ (119904

119894= 119904 (119905119904lowast 119894)) (2)

From this the sampling time 119905119904in turn is defined as

follows

119905119904= 119905119894+1

minus 119905119894 (3)

The quality of the data sampling toward zero Δ119889is

dependent on the accuracy of the function DQ which isusually implemented with ADCs or a comparator and theresolution of sampling time 119905

119904

The level-triggered interrupt-based sampling as shownin Figure 5(b) tries to capture the time-stampwhen it crossesa predefined condition with the parameterized value (such asthe voltage level) and its transition edge phase

Definition 2 Given continuous signal 119904(119905) let 119871 =

1198711 1198712 119871

119899 be a set of all levels of interest tomonitor data

from 119904(119905) TE a set of all pairs of the triggered level type andits elapsed time TE = te

119894| te119894= ⟨type et⟩ 119894 = 1 2 119899

and 119879119888119897119896

a fixed minimum period of the timer to measurethe time-stamp at a triggered point For the event te

119894isin TE

sampled time 119905119896is elapsed time after the previous event

te119894minus1

occurs and its quantization result error Δ119905by the

time-quantization function TQ is defined as follows

119905119896plusmn Δ119905= TQ (te

119894sdot et) = te

119894minus1sdot et + 119879

119888119897119896lowast 119896

(1 le 119896 lt infin)

(4)


Time

Mag

nitu

de

t0 t1 t2 timinus1 ti ti+1

s(t)

siminus1

si si+1Data sampling di

di ts

sn

middot middot middotmiddot middot middot

(a) Discrete-time sampling (active)

Time

Mag

nitu

de

Timely event

Time-sampling

Lm

teiminus1 teitei = ⟨Lm tk⟩

tk = teiminus1 middot et + Tclk lowast k

(b) Level-triggered interrupt-based sampling (passive)

Time

Mag

nitu

de

Event-quantizationElapsed

2nd detailed sample

1st wait

time et

aeviminus1 aevi = ⟨aeviminus1 dk Φedge tk⟩

tk = et + Tclk lowast k

Δd middot u lt |Lm minus dk| lt Δd middot

(c) Event capture approach by determining the presence of nextexpected atomic event in error range (active + passive)

ADCCMP

TMR OSC

Sensoranalog front-end

Atomic event data processingW

ait-t

ime

Onoff

On

off Trigger

sampling

Onoff

Repo

rtRe

port

Mai

n pr

oces

sor

AEG

Phas

e le

vel

Tracer memory aev

aevk

1st signal-to-eventconversion (S2E)

2nd detail

(d) Event-quantization based circuit data path

Figure 5 Event sampling illustrated and its circuit data path

The elapsed time for the 119894th triggered event is recursivelydetermined by searching the meet condition ldquo119896rdquo of thefollowing equation

te119894sdot et = te

119894minus1sdot et + 119905

119896

forallDQ (119871119898) = DQ (119904 (119905

119896)) 119871

119898isin 119871

(5)

The time-stamp te119894sdot time resolution toward zero Δ

119905is

dependent on the minimum value of the time advance (119879119888119897119896)

and the number-of-bits representation of (119896) value to encodethe time value Higher resolution of the time value requirescontinuous operations of the oscillator and timer unit withhigher accuracy and large size of a timer counter unit tomeasure the time-stamp which leads to energy consumptionoverhead as a side effect

Figure 5(c) describes our approach to capture the signalshape as an atomic event crossing a certain range of arrivaltime To more formally define our approach we begin ourexplanation by first presenting the following definitions

Definition 3 Given continuous signal 119904(119905) let AEV = aev119894|

aev119894= (aev

119894minus1 value phase et) be a sequence of an atomic

event aev119894crossing the specific level and time condition with

a relationship of previous atomic event aev119894minus1

where aev119894sdot

value is a result of the approximation-based data quantization

function ADQ and aev119894sdot et is a result of the approximation-

based time-quantization function ATQ described as follows

119889119896= ADQ (119904 (119905

119896) 119871119898 Δ119889 119906 V)

forallΔ119889lowast 119906 lt

1003816100381610038161003816119871119898 minus 119889119896

1003816100381610038161003816 lt Δ119889lowast V

119905119896= ATQ (aev

119894sdot et 119879119888119897119896)

where 119879119888119897119896

= 119879119888119897119896

+ Δ119905

(6)

The meet condition 119896 when the expected crossing ispresent is described in the following equation

119905119896= et + 119879

119888119897119896lowast 119896 forallDQ (119904 (119905

119896)) = 119889

119896 (7)

The atomic event generator (AEG) builds an element withthe attributes which are encoded with the digitized signallevel elapsed time and edge phase in the following equation

AEG (119904 (119905) 119871) = aev119894| aev119894= ⟨aev

119894minus1 119889119896 120601edge 119905119896⟩ (8)

From (8) the extracted information as an atomic eventis encoded with the approximation value of the signal levelthe reduced time-quantization value of the elapsed time andthe relationship of the previous atomic event aev

119896minus1

Figure 5(d) shows the proposed hardware data path forthe atomic-event sampling in Figure 5(c) including the level


comparators timer oscillator and atomic event generator(AEG) implementing operations from (2) to (8)

42 Atomic Event Segmentation Atomic event generation is amethod to represent a certain range of the continuous signalpattern with an abstract event The signal representationis classified by a user-defined set of signal segments Weprovide an example in Figure 6 The feature scan windowin Figure 6(a) which is used to capture the atomic event ofthe signal is configured with a specific voltage level timewindow and elapsed time at the feature pointThe configuredscan window determines if the featured points are monitoredin the snapshot of the signal passing through the configuredscan window and generates an element of a set of atomicevents in Figure 6(b)

Definition 4 Given the configured feature scanning windowto extract the atomic events from 119904(119905) let 119879start be a start timemonitoring the signal let 119879end be the end of monitoring thesignal let 119871

119903be a rising signal level at which the time-stamp

is 119879119903 let 119871

119891be a falling signal level at which the time-stamp

is 119879119891 let the pair of 119871

119909and 119879

119910be featured point and let119863max

be a maximum time value in which the featured points arepresent The set of signal segments described by the config-uration Ω = Ω

119894| Ω119894= (119879start 119879end 119871119903 119871119891 119879119903 119879119891 119863max)

of the featured scanning window is defined as Ω and is usedto extract the atomic events of interest for the AEG functionwhich is defined as follows

aev119894 = AEG (119904 (119905) Ω) (9)

Ωup defines a signal segment of the feature scanningwindow as an example showed in ldquoUp-Pulserdquo shape ofFigure 6(c) In our applications Ωtype | type = ldquouprdquo ldquosurdquoldquo sd rdquo ldquodprdquo ldquoisurdquo ldquoisdrdquo is used

Figure 6(c) shows examples of the user-defined signalshape as an atomic event which is determined by theconfigured feature scan windowThe ldquoup-pulserdquo pattern risesat recognized time 119877

119903and falls at recognized time 119877

119891for

the 119871probe level during maximum timer window 119863max Thecontinuous signal shape during a configured time range of119879start and 119879end is represented as abstract event 119877up with thefollowing equation

AEV (119904 (119905) Ωup) ≃ aevup = ⟨Ωup 119871probe (119877119903 119877119891)⟩ (10)

The expected rule to identify aevup atomic event patternallowing time error margin Δ is represented by the followingequation

119877up= (aevup Δ) (11)

The ldquostep-uprdquo pattern rises at time 119877119903and does not fall

within the timer window 119863max The continuous ldquostep-uprdquopattern signal can be also represented by the abstract atomicevent view in the following equation

AEV (119904 (119905) Ωsu) ≃ aevsu = ⟨Ωsu 119871probe (119877119903 minus)⟩ (12)

The infin-step-up pattern includes the specific range of nosignal transition crossing the specified voltage level 119871probewithin maximum timer window 119863max and a rise at anytime The continuous infin-step-up pattern signal can be alsorepresented by the abstract atomic event in the followingequation

AEV (119904 (119905) Ωisu)

≃ aevisu = ⟨Ωisu 119871probe (infininfin 119877119903)⟩

(13)

The aevisu atomic event pattern is also represented by theexpected atomic event pattern rule and its time error marginusing the following equation

119877isu

= (aevisu Δ) (14)

Theinfin-step-up pattern is a powerful method to simplifythe representation of the long-term signal shape with noactivity which leads to a reduction in the capacity of theinformation

Figure 6(d) shows the capability to represent various sig-nal shape by the configuration of 119871

119903119863max 119879119903 119879start and 119879end

in the feature scan window One signal shape can be dividedinto the several slices by user-defined signal segmentationIf the time window for signal segmentation is the same asthe fixed width 119905

119904in the discrete-time sample method the

result of the atomic event generation is equivalent to thatof the discrete-timed sampling The proposed atomic eventgeneration approach enables a trade-off between the signalextraction accuracy and its processing power consumption

43 Event-Driven Data Processing The atomic event genera-tor (AEG) scans the continuous signal 119904(119905) passing throughthe configured feature scan window to determine the pres-ence of the signal shapes of interest as shown in Figure 7(a)The set of atomic events is generated with a pair of attributesand time-stamps as a result of the time-quantization shownin Figure 7(b) Consider

aev = aev119894| aev0 aev1 aev

119894= (ldquo119871

119894rdquo 119905119904119894) (15)

Figure 7(c) shows a signal representation by a set ofatomic events with a certain amount of error This is denotedin the following equation

ae = ae119894| ae0 ae1 ae

119894= (ldquo119871

119894rdquo 119905119904119894plusmn Δ) (16)

aev119894 which is matched with the configured scan window

AE119894 is represented as an abstracted atomic event index in

Figures 7(d) and 7(e) which indirectly address the detailedattributes in the constant dictionary The continuous analogsignal is converted into a set of event quantized data aev

119894

and its index value is traced only into the atomic eventtracer buffer Therefore the traced event data processingmanipulates the index value and its relationship to the rep-resentative atomic events to generate the final event EV Theproposed event-driven sensor data processing unit (EPU)which is based on event-quantization provides the followingadvantages compared to conventional sensor data processing


Configuration of feature scanning window Ωi = (Lf Tr Dmax Tstart Tend )

Lf

Tr

Dmax

Tstart Tend

(a) Configuration of signal scanning window for atomic event extraction

Sensor analogsignal s(t)

Ω = Ωi set of signal segmentsE = aevi set of atomic events

AEGf Ω rarr E

Atomiceventaevi

(b) Atomic event generator definition

Time measurement window

Timer window

Timer startTimer end

Ti

Up-pulse (Rup) Step-up (Rsu)

Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)

Down-pulse (Rdp) infin

infin infin

-step-up (Risu ) infin-step-down (Risd)Rr

Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax

Rf gt DmaxRr gt Dmax

(c) Examples of set of signal segmentΩ

Elapsed time sweep of feature point

Swee

p of

feat

ured

leve

l

Equivalent to discrete-timed sampling

Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3

Tstart TstartTend Tend


⋱

(d) Representing various atomic events according to featured points

Figure 6 Configuration of signal feature scan window and atomic event segmentation


L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

(a) Event Driven Sampling

ae0 ae1 ae2 ae3 ae4ae5

ts0 ts1 ts2ts3 ts4 ts5

(b) Atomic Event Representation (time-quantization)

AE0 AE1 AE2 AE3 AE4 AE5

ts0plusmnΔ119905ts1plusmnΔ119905



(c) Allowing Accuracy Error


(d) Atomic Event Conversion Result mapping to pre-defined atomicevent (with index number)

Arrived atomic event

Expected atomic event

Quantized atomic event

ae0 = (ldquoL0rdquo ts0 )ae1 = (ldquoL1rdquo ts1 )ae2 = (ldquoXrdquo ts2 )ae3 = (ldquoL0rdquo ts3 )ae4 = (ldquoL1rdquo ts4 )ae5 = (ldquoL1rdquo ts5 )

AE0 = (ldquoL0rdquo ts0 plusmn Δt)AE1 = (ldquoL1rdquo ts1 plusmn Δt)AE2 = (ldquomdashrdquo ts2 plusmn Δt)AE3 = (ldquoL0rdquo ts3 plusmn Δt)AE3 = (ldquoL1rdquo ts4 plusmn Δt)AE3 = (ldquoL1rdquo ts5 plusmn Δt)

aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo

(e) Example value of event-quantization result

s(t)S2E

Tracer

AEG atomic event generatorCMP level comparatorTMR time-stamp timer

In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩

(f) Data-path of event-driven sensor data processing

Figure 7 Event-driven sensor data processing concept for macro-level signal analysis

(i) representing the continuous analog signal with asmall number of atomic events for specially featuredpoints

(ii) decreasing the number of pieces of processing datawith reduced atomic events

(iii) allowing the accuracy error of the generated atomicevents for application-specific properties of the rare-event sensing applications

(iv) decreasing the complexity of the expected atomicevent comparison circuit by comparing the time-stamp range instead of the accurate value

(v) mapping the recognized atomic events into the repre-sentative atomic event set with only index value

(vi) transforming the raw data processing into the indexvalue

Figure 7(f) illustrates the corresponding data path of theevent-driven sensor data processing including the atomic

event generator tracer feature scan window and the patternmatcher (which is described as event-print window matcherin Figure 8(b))

44 Final Event Identification The archived atomic events inthe tracer memory are evaluated as a similarity factor whichis calculated by the total sum of the distance between thecollected events and the expected rule We define this proce-dure as 119877lowast-plain projection as illustrated in Figure 8(a) Theatomic event extraction procedure is described by the 997888997888997888rarrAEVin the following equation The operation ⨂ describes theatomic event conversion for the continuous sensor signal 119904(119905)with the atomic event conversion rule which is introduced inFigure 6(c) Consider

119904 (119905)⨂1198710= AEV (17)


Traced atomic events (collected)

Long-term no activity Long-term no activity

Scan-window for atomic event conversion

Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6

Similarity factor 120582 = sumΔlowast

if 120582 le 120582min for detecting EVk

⊙

⊙

Rlowasti extract AEVlowast =

Rlowast-planeis identified as EVk

aev aev aev aev aev aev aev

aevi | aevi isin AEV for (aevi)0 == ldquolowastrdquosum (|(aevi) middot et minus

minusminusminusrarrAEVminusminusminusrarrAEV



i=01m

(Rlowasti ) middot et|) rarr Δlowast

(a) 119877-plain projection to represent similarity factor including accuracy error

Interest atomic event extraction

Determine whether the sequence of extracted atomic events ismatched with the expected rules

Final event identification using expected rule of atomic events

aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd

1 minusminusminusrarrAEV




(b) Accuracy-error-allowed pattern matcher implementation

Figure 8 Event-print identification for atomic event of interest

In the signal example of Figure 8(a) the result of atomicevent generation is described by the following equation

997888997888997888rarrAEV

= aevup0 aevsu1 aevsd2 aevisu3 aevsd4 aevsu5 aevisd6

(18)

Figure 8(b) shows the proposed event-print windowmatcher implementing the 119877lowast-plain projection to determinethe final event using the result based on the similarity factorThe event-print matcher allows a certain error comparingthe arrived atomic events to the expected rule which isillustrated as blank holes in the punch card of Figure 8(b)Multiple 119877lowast-plain projections are performed simultaneouslyto compare the archived atomic events with various patternrules We define these matching operations ⨀ with thefollowing equation

997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)

The extracted atomic event vector is described by thefollowing equation997888997888997888997888rarrAEVlowast

= aevlowast119894| aev119894isin AEV for (aev

119894) sdot type == ldquolowastrdquo

(20)

The difference between the elapsed time stamp value andexpected event arrival value is calculated for the extractedatomic events list using the following equation

Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)

The operation ⨀ of the extraction rule 119877lowast

119894for atomic

event vector 997888997888997888rarrAEV can be considered as the 119877lowast119894plane projec-

tion

The total similarity factor which is compared to theexpected event rules is described as the summation of thedifference in the measured time-stamp

120582 = sumΔlowast

119894 (22)

If the final value 120582 is less than the minimum value ofthe expected error range the received atomic event 997888997888997888rarrAEVis identified as EV

119896 The trade-off between the processing

accuracy and corresponding energy consumption can beselected to satisfy the design specification which is describedby the functional constraints and the required operatinglifetime

5 Implementation and Experimental Results

The proposed event-driven sensor data processing unit(EPU) is implemented as an accelerator to perform energy-efficient event recognition from the incoming sensor signalas described in Figure 9(a) The regular case for sensing dataanalysis can be covered by the proposed event processorwhich enables theMCU core to hibernate during sleepmodeThe user-defined software configured by the MCU coreallocates the configuration of the predefined atomic eventconversion conditions for the feature scan window

The set of atomic events is redirected into the tracer buffervia the dedicated DMA bus Atomic event generation andevent vector construction are performed in silent backgroundmode without waking up any of the main MCUs

The newly designed event-driven sensor processorincluding the general-purpose MCU core and the EPUcore is implemented using 018 120583m CMOS embedded-flashprocess technology and has a die size of 12mm times 12mmas shown in Figure 9(b) The proposed method requiresapproximately an additional 7500 logic gates using a 2-input


AddressDataInterrupt

MCU CPU core(general processing)

Port IO Timer subsystem Sensor analog

front-end

IRAM

Interrupt service routineCode ROM

(ISR)

Pattern matcherTracer

Configuration

Sensors

Irregular eventRegular event

Reporting the recognized events

Report

S2E converter

Signal-to-event conversion

Sensor IF

Tracer

Event matcher

Event generation

Recognized event

Internet connectivitiy

Monitored signal

(a) Proposed sensor processor architecture (b) Hierarchical event data processing

Tracer memory

Clock Timer

S2E

(c) Chip implementation

On-chip flash memory

(ISR code)

Sensor analog front-end

si Area overhead7500 gates (2-in NAND gates)

+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2

1kB SRAM trace (shared with stack memory)

1kB

Regi

sters

aevk998400

Figure 9 VLSI IC implementation of microcontrollers with the proposed event signal processing unit (EPU)

NAND for the timer counter signal-to-event converter (S2E)including AEG blocks event-print window matcher in EPUunit and 1-KB SRAM buffer for the atomic event vectortracer

Figure 10(a) shows the experimental design andmeasure-ment method to validate the efficiency in processing energyconsumption by the proposed method and its implementedhardware The first step of the evaluation is performed at thesimulation level using the MatlabSimulink models

The physical sensor signal acquisition by the real activity(such as gesture swipe proximity and human presence) isperformed off-line to save the raw dump file of time-variantsensor signal Then the raw file of the archived sensor signalis loaded into the Matlab workspace Figure 10(b) shows thatthe proposed event processing flow is evaluated by themodel-based designs using discrete-event design tool sets supportedby the Simulink

The second step of the evaluation is performed bythe circuit-level simulation for the synthesizable hardwaredesign The proposed method and its hardware architectureare implemented with the fully synthesizable Verilog RTLswhich can be physically mapped onto the FPGA or CMOSsilicon chip

The implemented chip includes the test mode interfacewhich requires about 1800 logic gates to access the on-chip registers and the bus transactions in supervisor modeIf the predefined test sequences are forced into the inputports on the power-up duration the chip is entered into thesupervisor mode in which the important nodes of the systemcan be accessed by the external interface The user-driven

external trigger events are loaded into the on-chip via the testmode interface to emulate the dynamic operation executinguser applications The event detection timing in chip-level iscompared to the expected timing and the power consumptionis alsomeasured in the real environment by the user-triggeredeventsThehardware overhead for the testmode interfacewillbe excluded in a final chip for the mass production

The third step compares the simulation results with theelectrical results which are measured for the implementedIoT sensor device including the proposed sensor processorBecause the gate-level synthesized design files are nearlyequivalent to the physical hardware the power simulationresults show that the proposed sensor processor architecturemay reduce energy consumption

The implemented IoT sensor device is a type of IR-(infrared radio-) based standalone system that senses achange in the movement of an object The IR transmitterIR receiver the proposed sensor processor Bluetooth for thewireless connectivity and on-board battery are integrated ona single tiny PCB board as shown in the board screenshot ofFigure 10(a)

The IR transmitter generates a specific pattern of signalpulses The IR receiver acquires the light signal reflected bythe target objects and the implemented sensor processorperforms signal processing to analyze the meaning of thesignal

Figure 11 shows the experimental results by the imple-mented IoT device comparing the operating lifetime accord-ing to the number of sample differences by the processingmethod in Figure 11(a) The S2D describes the result by the


Energy (power) accuracy error

simulation result

Power (energy) operating lifetime accuracy error

measurement result

Model-based simulation

Implementation measurement

Sensor signal raw data

Implemented system (IC)

Top HW design model


Circuit level simulation

Sensing application

Physically synthesized result

Physical power simulation result

Circuit-level operation verification

Implemented IoT deviceObject gesture recognition and indication system

via bluetooth (based on IR signal reflection)

Sensing application

Sensing application

Sensor signal raw data (

Implemented system (lowast

ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)

(a) Experimental design to compare with the implementation result

Figure 10 Continued


State flow

IR (infrared radio) signal reflection-based applications

(1) Gathered sensordata examples

(2) Loading into workspace ofMATLAB

(3) Signal-to-event(S2E) generation

(4) Atomic eventtracer

(5) Event dataprocessing and

matching

Proximity

Gesture swipe

Human presence

[middot [middotmiddot[middot [middotmiddot

Signal wave

From

Out Out

Outvc

vc

workspace FDA tool

Digitalfilter design

4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product

Constant 1V fallcrossing

Product11V falling edge Timed to

event signal2

Timed toevent signal1 Event-based

entity generator

Event-basedentity generator1

AE

AEIn

Out

Out

In

In2

In1

Set attribute

Set attribute1AE LEVEL

Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu

1V level falling edge

25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout

Instantaneous entitycounting scope1

Schedule timeout1 Enabled gate

Path combiner1 Single sever

In1

In2

Signal scope

Wait until 5 elements are tracedinto the buffer

AE level AE phase Instantaneous entitycounting scope

AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE

(b) MatlabSimulink-based Event-Driven Modeling amp Simulation

Figure 10 Experimental design and measurement of the implementation result

conventional polling-based sensor data processing methodThe S2T shows a result from the level-triggered interrupt-based signal processing method The S2E shows the resultsfromour proposedmethodThe improved results fromallow-ing an accuracy error are also evaluated These results whichallow for a 25 error variation in the specific constraints inour applications are shown in Figures 11(a) and 11(b)

All building blocks in the implemented IoT sensordevice hibernate during sleep mode except that of the EPU(including the signal-to-event converter) which is only activein order to trace the transition of the incoming signal bycomparing the signal features of interestWhen the final eventis identified by the event-print window matcher the mainMCU wakes up to activate the subsequent building blocksto transfer the recognized events to a host system using thewireless connectivity

As sensing applications of the implemented IoT sensordevice the low-power recognition performance based onthe proposed method is evaluated in terms of its energyconsumption Figure 11(c) shows the results for the specificIR sensor signal recognition using a defined set of signal

segments Ω The figure shows an 80 reduction when a 10accuracy error is allowed compared to the original work

Figure 11(d) shows the energy-efficient recognition of thehand-swipe gestures which are described by the two typesof signal segments Ω and show a 50 reduction when a10 accuracy error is allowed compared to the originalwork The reduction in energy consumption is achievedby configuring the implemented architectural frameworkwith application-specific constraints that allow the requiredrecognition accuracy error

Themargin of the acceptable accuracy error is dependenton the application-driven requirementThe trade-off betweenthe event detection accuracy and low-power consumptionhas to be considered in implementing the IoTdeviceThepro-posed chip architecture provides the configuration registerto allow the user-defined accuracy error for more long-termoperation of the IoT device In the environmentalmonitoringapplications such as proximity hand gesture temperatureand light intensity the response error in less than severalseconds for event detection is small enough to allow the 50accuracy error


5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)

times103 S2D versus S2T versus S2E (error sweep)

S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524

Difference of number of data samplesand number of events (16 1)

Operating lifetime comparison using battery capacity 1035mA h

(a) Operating lifetime according to the difference of the number ofsamples

S2D S2TEnergy 454 320 314 236 182 158Lifetime 7516 10683 10890 14453 18723 21685

0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)

Energy consumption and operating lifetime comparisonS2D versus S2T versus S2E (sample difference 1324) times103

S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s

(b) Operating energy amp lifetime comparison according to the accuracy

Event sampling

Up-pulse (Rup)

s(t)

Rup Rsd Rup Rsd Rup RupRsd

aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)

Processing 590 549 194 194 196 194Time measure 0 222 173 145 91 66Sampler 1392 24 12 12 12 12Lifetime 1714 4284 7947 9707 11456 12530

02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)

Energy consumption and operating lifetime comparisonS2D versus S2T versus S2E

times103times102

5x reduction(10 error)


Ope

ratin

g en

ergy

(120583J)

dur

ing1

s

(c) Energy consumption of proposed method for specific IR reflection signal pattern recognition

S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei

Step-down (Rsd)Ω Step-up (Rsu)

s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s

(d) Energy consumption of proposed method for IR signal reflection arrival time distance measurement (Gesture Swipe Applications)

Figure 11 Experimental results


6 Conclusion

The proposed event-driven sensor processor architecture forsensing applications with rare-event constraints is proposedand implemented as an accelerator This enables the sensorsignal processing in an energy-efficient mode by allowingfor the accuracy error which is caused by the abstractionof the original signal as atomic events All of the buildingblocks including atomic event generators and the EPU coreare implemented as a single system-on-a-chip which isintegrated into the IoT sensor device

The proposed method uses the characteristics of rare-event sensing applications in which timing accuracy error isrelatively insensitive in sensor signal recognition and intro-duces the concept of atomic event generation as a methodof event-quantization The event-space representation ofthe sensor signal by the extracted set of atomic events isconstructed with a user-defined event signal segmentation bythe configured feature scan window

The event-quantization result is archived as a set ofunique indexes into the tracer bufferThe event-printmatcherdetermines the presence of the featured signal points for thecollected atomic event vector in order to identify the eventfrom the original sensor signal The event-matching processis based on the similarity factor calculation named 119877lowast-plainprojection that describes the expected rules of the signalpatterns

The implementation results which are evaluated foran IR-based signal recognition system for object activitymonitoring applications show a reduction in total energyconsumption by delaying the activation of themain processorand the Bluetooth interface for the wireless connectivityThe proposed sensor processor provides an architecturalframework by providing an application-specific configura-tion of the event-quantization level for the energy-accuracytrade-off This results in additional benefit of the energyconsumption which maximizes the operating lifetime of theIoT sensor device

Conflict of Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea (NRF) funded by the Ministry of Education(2014R1A6A3A04059410) the BK21 Plus project funded bytheMinistry of Education School of Electronics EngineeringKyungpook National University Korea (21A20131600011)and the MSIP (Ministry of Science ICT amp Future Planning)Korea under the C-ITRC (Convergence Information Tech-nology Research Center) support program (NIPA-2014-H0401-14-1004) supervised by the NIPA (National ITIndustry Promotion Agency)

References

[1] H T Chaminda V Klyuev and K Naruse ldquoA smart remindersystem for complex human activitiesrdquo in Proceedings of the14th International Conference on Advanced CommunicationTechnology (ICACT rsquo12) pp 235ndash240 2012

[2] L Chen J Hoey C D Nugent D J Cook and Z Yu ldquoSensor-based activity recognitionrdquo IEEE Transactions on Systems Manand Cybernetics C Applications and Reviews vol 42 no 6 pp790ndash808 2012

[3] K Choi R Soma and M Pedram ldquoFine-grained dynamicvoltage and frequency scaling for precise energy and perfor-mance tradeoff based on the ratio of off-chip access to on-chip computation timesrdquo IEEETransactions onComputer-AidedDesign of Integrated Circuits and Systems vol 24 no 1 pp 18ndash28 2005

[4] A B da Cunha and D C da Silva Jr ldquoAn approach for thereduction of power consumption in sensor nodes of wirelesssensor networks case analysis of mica2rdquo in Proceedings of the6th International Conference on Embedded Computer SystemsArchitectures Modeling and Simulation (SAMOS 06) vol 4017of Lecture Notes in Computer Science pp 132ndash141 SpringerSamos Greece July 2006

[5] B French D P Siewiorek A Smailagic and M DeisherldquoSelective sampling strategies to conserve power in contextaware devicesrdquo in Proceedings of the 11th IEEE InternationalSymposium on Wearable Computers (ISWC rsquo07) pp 77ndash80Boston Mass USA October 2007

[6] V Gupta D Mohapatra A Raghunathan and K Roy ldquoLow-power digital signal processing using approximate addersrdquoIEEE Transactions on Computer-Aided Design of IntegratedCircuits and Systems vol 32 no 1 pp 124ndash137 2013

[7] M Hempstead N Tripathi P Mauro G-YWei and D BrooksldquoAn ultra low power system architecture for sensor networkapplicationsrdquoACMSIGARCHComputer Architecture News vol33 no 2 pp 208ndash219 2005

[8] M Hempstead G-Y Wei and D Brooks ldquoArchitecture andcircuit techniques for low-throughput energy-constrained sys-tems across technology generationsrdquo in Proceedings of the Inter-national Conference on Compilers Architecture and Synthesis forEmbedded Systems (CASES rsquo06) pp 368ndash378 New York NYUSA October 2006

[9] A B Kahng and K Seokhyeong ldquoAccuracy-configurable adderfor approximate arithmetic designsrdquo in Proceedings of theACMEDACIEEE 49th Annual Design Automation Conference(DAC rsquo12) pp 820ndash825 ACM San Francisco Calif USA June2012

[10] K van Laerhoven H-W Gellersen and Y G Malliaris ldquoLong-term activity monitoring with a wearable sensor noderdquo inProceedings of the International Workshop on Wearable andImplantable Body Sensor Networks (BSN rsquo06) pp 171ndash174 April2006

[11] O D Lara and M A Labrador ldquoA survey on human activityrecognition using wearable sensorsrdquo IEEE CommunicationsSurveys and Tutorials vol 15 no 3 pp 1192ndash1209 2013

[12] M T Lazarescu ldquoDesign of a WSN platform for long-termenvironmental monitoring for IoT applicationsrdquo IEEE Journalon Emerging and Selected Topics in Circuits and Systems vol 3no 1 pp 45ndash54 2013

[13] S-W Lee and K Mase ldquoActivity and location recognition usingwearable sensorsrdquo IEEE Pervasive Computing vol 1 no 3 pp24ndash32 2002


[14] Y Lee J Kim and C-M Kyung ldquoEnergy-aware video encodingfor image quality improvement in battery-operated surveillancecamerardquo IEEE Transactions on Very Large Scale Integration(VLSI) Systems vol 20 no 2 pp 310ndash318 2012

[15] K Leuenberger and R Gassert ldquoLow-power sensor module forlong-term activity monitoringrdquo in Proceedings of the AnnualInternational Conference of the IEEE Engineering in Medicineand Biology Society (EMBC rsquo11) pp 2237ndash2241 2011

[16] W Li M Bandai and T Watanabe ldquoTradeoffs among delayenergy and accuracy of partial data aggregation in wirelesssensor networksrdquo in Proceedings of the 24th IEEE InternationalConference on Advanced Information Networking and Applica-tions (AINA rsquo10) pp 917ndash924 April 2010

[17] M Miskowicz ldquoThe event-triggered integral criterion forsensor samplingrdquo in Proceedings of the IEEE InternationalSymposium on Industrial Electronics (ISIE rsquo05) vol 3 pp 1061ndash1066 2005

[18] P Panek ldquoError analysis and bounds in time delay estimationrdquoIEEE Transactions on Signal Processing vol 55 no 7 pp 3547ndash3549 2007

[19] D Park C M Kim S Kwak and T G Kim ldquoA low-powerfractional-order synchronizer for syncless time-sequential syn-chronization of 3-D TV active shutter glassesrdquo IEEE Transac-tions on Circuits and Systems for Video Technology vol 23 no2 pp 364ndash369 2013

[20] B Schell and Y Tsividis ldquoAnalysis of continuous-time digitalsignal processorsrdquo in Proceedings of the IEEE InternationalSymposium on Circuits and Systems (ISCAS rsquo07) pp 2232ndash22352007

[21] J Sorber A BalasubramanianMD Corner J R Ennen andCQualls ldquoTula balancing energy for sensing and communicationin a perpetual mobile systemrdquo IEEE Transactions on MobileComputing vol 12 no 4 pp 804ndash816 2013

[22] Y Tsividis ldquoEvent-driven data acquisition and continuous-timedigital signal processingrdquo in Proceedings of the IEEE CustomIntegrated Circuits Conference (CICC rsquo10) pp 1ndash8 2010

[23] Y Tsividis ldquoEvent-driven data acquisition and digital signalprocessingmdasha tutorialrdquo IEEE Transactions on Circuits andSystems II Express Briefs vol 57 no 8 pp 577ndash581 2010

[24] Y Yilmaz GMoustakides and XWang ldquoSpectrum sensing viaevent-triggered samplingrdquo in Proceedings of the 45th AsilomarConference on Signals Systems and Computers (ASILOMAR rsquo11)pp 1420ndash1424 November 2011

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



Power

Sensor

Sampling (ADC CMP)

CPU execution

Voltage

Long-term

Long-term

no activity

Triggered event

Internet connection

Recognized

activity

PsensorPsampling

Pup

Tduty

Tsensor Tsampling

Tup

TwakeupTCPU

Tinternet

2mA

16mA

600120583A

350120583A

t

t

Curr

entlowast

5V

Sensor signal s(t)

Atomic event (aevi)

event (evk)

aev0 aevi

ev0

ev1

Pleakage

Figure 1 Power consumption for sensor signal sampling andprocessing in long-term activity monitoring

even during the period of long-term sleep Although thisweakness in terms of the power consumption can be resolvedusing several programming techniques where the interrupthandlers consider the sensor signal behavior this is notformal approach that covers the general cases of signalcharacteristics

Several approaches [6 9ndash11] regarding the behavior ofsensing applications have been introduced in sensor-basedapplication for low power operations especially in activitysensor sensor processor andwireless sensor nodes Howeverthe use of the discrete-time-based operation is restrictedbecause the most feasible devices rely on commercial off theshelf (COTS) MCUs attached with several sensors Thesedevices attempt to control the ratio of the sleep mode tothe active mode by switching the operation mode to lowerthe power consumption according to being with or withoutthe sensory data which are continuously determined by thediscrete-time-based monitoring of the signal

2 Motivation

To address this limitation of the conventional digital sys-tem architecture by using the discrete-time-based sen-sor data processing method we propose an event-drivensystem architecture that modifies traditional digital systemdesign We present a theoretical framework to implement anevent-driven sensor processor for general rare-event sensingapplications by analyzing the system operations

21 Event-Space Signal Representation Our main researchbegins with an event-space representation of the sig-nal instead of the digital data space domain Figure 2(a)

illustrates fundamental signal elements using attributesof interest and Figure 2(b) shows the event data spacetransition-based representation which describes in detail therelationship between the featured points of the sensed signal

The extracted features of the sensed signal are encodedinto the elapsed time between events and informative valuesuch as voltage level and edge phase crossing the trigger pointof the signalThe fundamental event defined which is definedas an atomic event with the most important informationprovides a signal representation on an abstract level andreduces the computational complexity in performing basicdata processing for extracted informative features of interestThe collected atomic events include partial information inthe original signal that specifies whether the desired featuredpoints of the signal are present

22 Event-Quantization with Accuracy Error The event-quantization concept extends the time-quantization methodfor signal representation that uses elapsed time to enhance theconventional data sampling and processing method Time-quantization monitors only the specific conditions of thesignal transition and captures the time-stamps The event-quantization method also determines whether the specifiedcharacteristics of the signal exist Figures 2(c) and 2(d)illustrate the difference between the time-sampling andevent-quantization methods with an accuracy error whichmonitors only the presence of the featured points of interest

In terms of accuracy there are two types of operationstiming accuracy and data resolution accuracy [12] Theformer is dependent on the sampling frequency comparedto the received signal bandwidth and the latter is derivedfrom the representative resolution of the sampled data Thetiming resolution is dependent on the clock-duty resolutionto resolve the timing window of the processing operationsaccurately

Higher timing resolution for the duration measurementrequires the accurate data processing in the time domainHowever inmost cases the data value resolution of the targetsystem is tightly required but timemdashrelated specificationincluding the response time is relatively allowable in a certainrange of error In particular in rare-event sensing applica-tions we assume that the timing measurement accuracy forthe elapsed time between the arrived signal events can beperformed with inaccurate operating clock relatively

Although the events contain inaccurate time informationthe final event receiver (for example a human interface) isunable to identify the error differences compared to the idealdata within a certain error range [13]

23 Accuracy Configurable Signal-to-Event Conversion Theevent-based approach with a certain amount of accuracyerror is described by the proposed event-driven sensor dataprocessing flow as shown in Figure 2(e) The input signalis monitored with specified interest-of-signal characteristics119882(119896) to generate the specific atomic events aev

119894of the signal

The set of atomic events during the specified region of thesignal are traced into the tracer memory as an event vector997888997888997888rarrAEV which contains the sequence of the atomic events and




Distance

s(t)

s(t)

thc

thb

tha




t

Sens

or si

gnal

sp

ace

t

Ea

Eb

Ec

EdEven

t dat

a spa

ce

= ⟨Ec et4⟩aev4

= ⟨Ed et5⟩aev5





High accurate clock



s(t)

OSC

tdi

(thk eti)



OSC

aevi



gttimes

times

Δe

z(t)

z(k)

Event tracing


(threshold levels)

Atomic event


event vector




Cs

Ds

EVj

Z(k)|k=⟨sel⟩

= (eyi et i)

= (Ei eti)

Δe



minusminusrarrRV








Time


Dk = Δs middot k

ti = ts lowast i

s(t)


Event-quantization

OSC

Time

Data

s(t)

Tk = Δt middot k

li = ls lowast i

⟨li Tk⟩


Time

Data

Time-quantization

OSC

s(t)

Tk = Δt middot k

li = ls lowast i


(a) Target


Data

Dk = Δs middot k

s(t)



Time

Data


quantization


s(t)




Time

Data



s(t)



(b) Accuracy

t

t

t

OSC

t

tEvent-

capturedEvent-

triggered


s(t)tha

aeidealae0

dclk

Δosc Δev

eclk

ΔaΔe


ae0998400

(c) Accuracy error










em998400 m

thi = Δ lowast i

thi = Δ lowast i

thi = Δ lowast i

n gt m n asymp m



s(t)

s(t)

s(t)

Time

Time



Time








1198981015840119898


119889119898119896

≫ 1198901198981015840119898 (1)








119899 a uniformly


119899 A sampled


119889by


119889119894plusmn Δ119889= DQ (119904

119894= 119904 (119905119904lowast 119894)) (2)


follows

119905119904= 119905119894+1

minus 119905119894 (3)



119904



1198711 1198712 119871



119894| te119894= ⟨type et⟩ 119894 = 1 2 119899

and 119879119888119897119896


119894isin TE


te119894minus1



119905119896plusmn Δ119905= TQ (te

119894sdot et) = te

119894minus1sdot et + 119879

119888119897119896lowast 119896


(4)


Time

Mag

nitu

de


s(t)

siminus1


di ts

sn



Time

Mag

nitu

de

Timely event

Time-sampling

Lm




Time

Mag

nitu

de


2nd detailed sample

1st wait

time et





ADCCMP

TMR OSC



ait-t

ime

Onoff

On

off Trigger

sampling

Onoff

Repo

rtRe

port

Mai

n pr

oces

sor

AEG

Phas

e le

vel

Tracer memory aev

aevk


2nd detail





119894minus1sdot et + 119905

119896

forallDQ (119871119898) = DQ (119904 (119905

119896)) 119871

119898isin 119871

(5)


119905is





aev119894= (aev




where aev119894sdot




119889119896= ADQ (119904 (119905

119896) 119871119898 Δ119889 119906 V)


1003816100381610038161003816119871119898 minus 119889119896

1003816100381610038161003816 lt Δ119889lowast V

119905119896= ATQ (aev

119894sdot et 119879119888119897119896)

where 119879119888119897119896

= 119879119888119897119896

+ Δ119905

(6)


119905119896= et + 119879

119888119897119896lowast 119896 forallDQ (119904 (119905

119896)) = 119889

119896 (7)


AEG (119904 (119905) 119871) = aev119894| aev119894= ⟨aev

119894minus1 119889119896 120601edge 119905119896⟩ (8)


119896minus1







is 119879119903 let 119871


is 119879119891 let the pair of 119871

119909and 119879



119894| Ω119894= (119879start 119879end 119871119903 119871119891 119879119903 119879119891 119863max)


aev119894 = AEG (119904 (119905) Ω) (9)




119891for




119877up= (aevup Δ) (11)





AEV (119904 (119905) Ωisu)


(13)


119877isu

= (aevisu Δ) (14)



119903119863max 119879119903 119879start and 119879end






119894= (ldquo119871

119894rdquo 119905119904119894) (15)



119894= (ldquo119871

119894rdquo 119905119904119894plusmn Δ) (16)




119894




Lf

Tr

Dmax

Tstart Tend




AEGf Ω rarr E

Atomiceventaevi



Timer window


Ti


Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)


infin infin


Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax




Swee

p of

feat

ured

leve

l


Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3



⋱




L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












Distance

s(t)

s(t)

thc

thb

tha




t

Sens

or si

gnal

sp

ace

t

Ea

Eb

Ec

EdEven

t dat

a spa

ce

= ⟨Ec et4⟩aev4

= ⟨Ed et5⟩aev5





High accurate clock



s(t)

OSC

tdi

(thk eti)



OSC

aevi



gttimes

times

Δe

z(t)

z(k)

Event tracing


(threshold levels)

Atomic event


event vector




Cs

Ds

EVj

Z(k)|k=⟨sel⟩

= (eyi et i)

= (Ei eti)

Δe



minusminusrarrRV








Time


Dk = Δs middot k

ti = ts lowast i

s(t)


Event-quantization

OSC

Time

Data

s(t)

Tk = Δt middot k

li = ls lowast i

⟨li Tk⟩


Time

Data

Time-quantization

OSC

s(t)

Tk = Δt middot k

li = ls lowast i


(a) Target


Data

Dk = Δs middot k

s(t)



Time

Data


quantization


s(t)




Time

Data



s(t)



(b) Accuracy

t

t

t

OSC

t

tEvent-

capturedEvent-

triggered


s(t)tha

aeidealae0

dclk

Δosc Δev

eclk

ΔaΔe


ae0998400

(c) Accuracy error










em998400 m

thi = Δ lowast i

thi = Δ lowast i

thi = Δ lowast i

n gt m n asymp m



s(t)

s(t)

s(t)

Time

Time



Time








1198981015840119898


119889119898119896

≫ 1198901198981015840119898 (1)








119899 a uniformly


119899 A sampled


119889by


119889119894plusmn Δ119889= DQ (119904

119894= 119904 (119905119904lowast 119894)) (2)


follows

119905119904= 119905119894+1

minus 119905119894 (3)



119904



1198711 1198712 119871



119894| te119894= ⟨type et⟩ 119894 = 1 2 119899

and 119879119888119897119896


119894isin TE


te119894minus1



119905119896plusmn Δ119905= TQ (te

119894sdot et) = te

119894minus1sdot et + 119879

119888119897119896lowast 119896


(4)


Time

Mag

nitu

de


s(t)

siminus1


di ts

sn



Time

Mag

nitu

de

Timely event

Time-sampling

Lm




Time

Mag

nitu

de


2nd detailed sample

1st wait

time et





ADCCMP

TMR OSC



ait-t

ime

Onoff

On

off Trigger

sampling

Onoff

Repo

rtRe

port

Mai

n pr

oces

sor

AEG

Phas

e le

vel

Tracer memory aev

aevk


2nd detail





119894minus1sdot et + 119905

119896

forallDQ (119871119898) = DQ (119904 (119905

119896)) 119871

119898isin 119871

(5)


119905is





aev119894= (aev




where aev119894sdot




119889119896= ADQ (119904 (119905

119896) 119871119898 Δ119889 119906 V)


1003816100381610038161003816119871119898 minus 119889119896

1003816100381610038161003816 lt Δ119889lowast V

119905119896= ATQ (aev

119894sdot et 119879119888119897119896)

where 119879119888119897119896

= 119879119888119897119896

+ Δ119905

(6)


119905119896= et + 119879

119888119897119896lowast 119896 forallDQ (119904 (119905

119896)) = 119889

119896 (7)


AEG (119904 (119905) 119871) = aev119894| aev119894= ⟨aev

119894minus1 119889119896 120601edge 119905119896⟩ (8)


119896minus1







is 119879119903 let 119871


is 119879119891 let the pair of 119871

119909and 119879



119894| Ω119894= (119879start 119879end 119871119903 119871119891 119879119903 119879119891 119863max)


aev119894 = AEG (119904 (119905) Ω) (9)




119891for




119877up= (aevup Δ) (11)





AEV (119904 (119905) Ωisu)


(13)


119877isu

= (aevisu Δ) (14)



119903119863max 119879119903 119879start and 119879end






119894= (ldquo119871

119894rdquo 119905119904119894) (15)



119894= (ldquo119871

119894rdquo 119905119904119894plusmn Δ) (16)




119894




Lf

Tr

Dmax

Tstart Tend




AEGf Ω rarr E

Atomiceventaevi



Timer window


Ti


Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)


infin infin


Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax




Swee

p of

feat

ured

leve

l


Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3



⋱




L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Time


Dk = Δs middot k

ti = ts lowast i

s(t)


Event-quantization

OSC

Time

Data

s(t)

Tk = Δt middot k

li = ls lowast i

⟨li Tk⟩


Time

Data

Time-quantization

OSC

s(t)

Tk = Δt middot k

li = ls lowast i


(a) Target


Data

Dk = Δs middot k

s(t)



Time

Data


quantization


s(t)




Time

Data



s(t)



(b) Accuracy

t

t

t

OSC

t

tEvent-

capturedEvent-

triggered


s(t)tha

aeidealae0

dclk

Δosc Δev

eclk

ΔaΔe


ae0998400

(c) Accuracy error










em998400 m

thi = Δ lowast i

thi = Δ lowast i

thi = Δ lowast i

n gt m n asymp m



s(t)

s(t)

s(t)

Time

Time



Time








1198981015840119898


119889119898119896

≫ 1198901198981015840119898 (1)








119899 a uniformly


119899 A sampled


119889by


119889119894plusmn Δ119889= DQ (119904

119894= 119904 (119905119904lowast 119894)) (2)


follows

119905119904= 119905119894+1

minus 119905119894 (3)



119904



1198711 1198712 119871



119894| te119894= ⟨type et⟩ 119894 = 1 2 119899

and 119879119888119897119896


119894isin TE


te119894minus1



119905119896plusmn Δ119905= TQ (te

119894sdot et) = te

119894minus1sdot et + 119879

119888119897119896lowast 119896


(4)


Time

Mag

nitu

de


s(t)

siminus1


di ts

sn



Time

Mag

nitu

de

Timely event

Time-sampling

Lm




Time

Mag

nitu

de


2nd detailed sample

1st wait

time et





ADCCMP

TMR OSC



ait-t

ime

Onoff

On

off Trigger

sampling

Onoff

Repo

rtRe

port

Mai

n pr

oces

sor

AEG

Phas

e le

vel

Tracer memory aev

aevk


2nd detail





119894minus1sdot et + 119905

119896

forallDQ (119871119898) = DQ (119904 (119905

119896)) 119871

119898isin 119871

(5)


119905is





aev119894= (aev




where aev119894sdot




119889119896= ADQ (119904 (119905

119896) 119871119898 Δ119889 119906 V)


1003816100381610038161003816119871119898 minus 119889119896

1003816100381610038161003816 lt Δ119889lowast V

119905119896= ATQ (aev

119894sdot et 119879119888119897119896)

where 119879119888119897119896

= 119879119888119897119896

+ Δ119905

(6)


119905119896= et + 119879

119888119897119896lowast 119896 forallDQ (119904 (119905

119896)) = 119889

119896 (7)


AEG (119904 (119905) 119871) = aev119894| aev119894= ⟨aev

119894minus1 119889119896 120601edge 119905119896⟩ (8)


119896minus1







is 119879119903 let 119871


is 119879119891 let the pair of 119871

119909and 119879



119894| Ω119894= (119879start 119879end 119871119903 119871119891 119879119903 119879119891 119863max)


aev119894 = AEG (119904 (119905) Ω) (9)




119891for




119877up= (aevup Δ) (11)





AEV (119904 (119905) Ωisu)


(13)


119877isu

= (aevisu Δ) (14)



119903119863max 119879119903 119879start and 119879end






119894= (ldquo119871

119894rdquo 119905119904119894) (15)



119894= (ldquo119871

119894rdquo 119905119904119894plusmn Δ) (16)




119894




Lf

Tr

Dmax

Tstart Tend




AEGf Ω rarr E

Atomiceventaevi



Timer window


Ti


Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)


infin infin


Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax




Swee

p of

feat

ured

leve

l


Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3



⋱




L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











em998400 m

thi = Δ lowast i

thi = Δ lowast i

thi = Δ lowast i

n gt m n asymp m



s(t)

s(t)

s(t)

Time

Time



Time








1198981015840119898


119889119898119896

≫ 1198901198981015840119898 (1)








119899 a uniformly


119899 A sampled


119889by


119889119894plusmn Δ119889= DQ (119904

119894= 119904 (119905119904lowast 119894)) (2)


follows

119905119904= 119905119894+1

minus 119905119894 (3)



119904



1198711 1198712 119871



119894| te119894= ⟨type et⟩ 119894 = 1 2 119899

and 119879119888119897119896


119894isin TE


te119894minus1



119905119896plusmn Δ119905= TQ (te

119894sdot et) = te

119894minus1sdot et + 119879

119888119897119896lowast 119896


(4)


Time

Mag

nitu

de


s(t)

siminus1


di ts

sn



Time

Mag

nitu

de

Timely event

Time-sampling

Lm




Time

Mag

nitu

de


2nd detailed sample

1st wait

time et





ADCCMP

TMR OSC



ait-t

ime

Onoff

On

off Trigger

sampling

Onoff

Repo

rtRe

port

Mai

n pr

oces

sor

AEG

Phas

e le

vel

Tracer memory aev

aevk


2nd detail





119894minus1sdot et + 119905

119896

forallDQ (119871119898) = DQ (119904 (119905

119896)) 119871

119898isin 119871

(5)


119905is





aev119894= (aev




where aev119894sdot




119889119896= ADQ (119904 (119905

119896) 119871119898 Δ119889 119906 V)


1003816100381610038161003816119871119898 minus 119889119896

1003816100381610038161003816 lt Δ119889lowast V

119905119896= ATQ (aev

119894sdot et 119879119888119897119896)

where 119879119888119897119896

= 119879119888119897119896

+ Δ119905

(6)


119905119896= et + 119879

119888119897119896lowast 119896 forallDQ (119904 (119905

119896)) = 119889

119896 (7)


AEG (119904 (119905) 119871) = aev119894| aev119894= ⟨aev

119894minus1 119889119896 120601edge 119905119896⟩ (8)


119896minus1







is 119879119903 let 119871


is 119879119891 let the pair of 119871

119909and 119879



119894| Ω119894= (119879start 119879end 119871119903 119871119891 119879119903 119879119891 119863max)


aev119894 = AEG (119904 (119905) Ω) (9)




119891for




119877up= (aevup Δ) (11)





AEV (119904 (119905) Ωisu)


(13)


119877isu

= (aevisu Δ) (14)



119903119863max 119879119903 119879start and 119879end






119894= (ldquo119871

119894rdquo 119905119904119894) (15)



119894= (ldquo119871

119894rdquo 119905119904119894plusmn Δ) (16)




119894




Lf

Tr

Dmax

Tstart Tend




AEGf Ω rarr E

Atomiceventaevi



Timer window


Ti


Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)


infin infin


Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax




Swee

p of

feat

ured

leve

l


Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3



⋱




L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Time

Mag

nitu

de


s(t)

siminus1


di ts

sn



Time

Mag

nitu

de

Timely event

Time-sampling

Lm




Time

Mag

nitu

de


2nd detailed sample

1st wait

time et





ADCCMP

TMR OSC



ait-t

ime

Onoff

On

off Trigger

sampling

Onoff

Repo

rtRe

port

Mai

n pr

oces

sor

AEG

Phas

e le

vel

Tracer memory aev

aevk


2nd detail





119894minus1sdot et + 119905

119896

forallDQ (119871119898) = DQ (119904 (119905

119896)) 119871

119898isin 119871

(5)


119905is





aev119894= (aev




where aev119894sdot




119889119896= ADQ (119904 (119905

119896) 119871119898 Δ119889 119906 V)


1003816100381610038161003816119871119898 minus 119889119896

1003816100381610038161003816 lt Δ119889lowast V

119905119896= ATQ (aev

119894sdot et 119879119888119897119896)

where 119879119888119897119896

= 119879119888119897119896

+ Δ119905

(6)


119905119896= et + 119879

119888119897119896lowast 119896 forallDQ (119904 (119905

119896)) = 119889

119896 (7)


AEG (119904 (119905) 119871) = aev119894| aev119894= ⟨aev

119894minus1 119889119896 120601edge 119905119896⟩ (8)


119896minus1







is 119879119903 let 119871


is 119879119891 let the pair of 119871

119909and 119879



119894| Ω119894= (119879start 119879end 119871119903 119871119891 119879119903 119879119891 119863max)


aev119894 = AEG (119904 (119905) Ω) (9)




119891for




119877up= (aevup Δ) (11)





AEV (119904 (119905) Ωisu)


(13)


119877isu

= (aevisu Δ) (14)



119903119863max 119879119903 119879start and 119879end






119894= (ldquo119871

119894rdquo 119905119904119894) (15)



119894= (ldquo119871

119894rdquo 119905119904119894plusmn Δ) (16)




119894




Lf

Tr

Dmax

Tstart Tend




AEGf Ω rarr E

Atomiceventaevi



Timer window


Ti


Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)


infin infin


Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax




Swee

p of

feat

ured

leve

l


Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3



⋱




L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation














is 119879119903 let 119871


is 119879119891 let the pair of 119871

119909and 119879



119894| Ω119894= (119879start 119879end 119871119903 119871119891 119879119903 119879119891 119863max)


aev119894 = AEG (119904 (119905) Ω) (9)




119891for




119877up= (aevup Δ) (11)





AEV (119904 (119905) Ωisu)


(13)


119877isu

= (aevisu Δ) (14)



119903119863max 119879119903 119879start and 119879end






119894= (ldquo119871

119894rdquo 119905119904119894) (15)



119894= (ldquo119871

119894rdquo 119905119904119894plusmn Δ) (16)




119894




Lf

Tr

Dmax

Tstart Tend




AEGf Ω rarr E

Atomiceventaevi



Timer window


Ti


Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)


infin infin


Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax




Swee

p of

feat

ured

leve

l


Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3



⋱




L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











Lf

Tr

Dmax

Tstart Tend




AEGf Ω rarr E

Atomiceventaevi



Timer window


Ti


Lprobe Lprobe

Lprobe Lprobe

Lprobe

Lprobe

Step-down (Rsd)


infin infin


Rr

RrRf

Rf

RfRf gt Dmax

Rf gt Dmax

Dmax

Rr gt Dmax

Rr gt Dmax




Swee

p of

feat

ured

leve

l


Signal segmentation

Lr1

Lr1

Lr1

Lr1

Lr1

Lr1

Tr1

Tr1

Tr1

Tr1

Tr2

Tr2

Tr3

Tr3



⋱




L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










L0

L1

s(t)

ae0

ae1 ae2 ae3 ae4 ae5

















aeaeaeaeaeae

0 = ldquoardquo

1 = ldquobrdquo

2 = ldquocrdquo

3 = ldquoardquo

4 = ldquobrdquo

5 = ldquobrdquo


s(t)S2E

Tracer


In

Out

CMP

AEG

OSC TMR

Li

MatcherEVx

ae0

ae1

ae2

aeiae5

AE0

AE1

AE2

AE5

ae1

ae2

ae5

ae0

⟨ ⟩

⟨ ⟩

⟨ ⟩

⟨ ⟩












119904 (119905)⨂1198710= AEV (17)





Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













Projection

L0

s(t)

up0

su1

sd2

isu3

sd4

su5

isd6

= aeup0 aesu

1 aesd2 aeisu

3 aesd4 aesu

5 aeisd6



⊙

⊙








i=01m






aevup0

aevsu1 aevsu

5

aevsd2 aevsd

4

aevisu3 aevisd

6

up0 ⊙ R

su0 Rsu

1 ⊙ Rsd0 Rsd

1 ⊙ Risu0 Risd

1 ⊙ R

up0 R

Rsu0 Rsu

1

Rsd0 Rsd

1

Risu1 Risd








997888997888997888rarrAEV


(18)


997888997888997888rarrAEV⨀119877lowast

119894119894=01119898

(19)




(20)


Δlowast

119894=1003816100381610038161003816(aevlowast

119894) sdot et minus (119877

lowast

119894) sdot et1003816100381610038161003816 (21)


119894for atomic


tion



119894 (22)












front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation













front-end

IRAM


(ISR)


Configuration

Sensors



Report

S2E converter


Sensor IF

Tracer

Event matcher

Event generation

Recognized event


Monitored signal


Tracer memory

Clock Timer

S2E



(ISR code)



+

ESPcore

tracerMCU + ESP

MCU +

aevj998400

aev0 aev1 aev2


1kB

Regi

sters

aevk998400














simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











simulation result


measurement result





Top HW design model



Sensing application






Sensing application

Sensing application



ΔΩ

1st step

2nd step

3rd step

Δe

(lowast middot c)

(lowast middot c)

(lowast middot c)

(lowast middot m)

lowast middot vcd)

lowast middot v)


Figure 10 Continued


State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










State flow







matching

Proximity

Gesture swipe

Human presence


Signal wave

From

Out Out

Outvc

vc

workspace FDA tool


4times

times

times25

++

ts1

5V rise

5V rise edge

crossing

Constant1 Product2

Product



event signal2


entity generator


AE

AEIn

Out

Out

In

In2

In1

Set attribute


Path combiner

= 1

AE LEVEL = 5

AE PHASE = minus1

AE PHASE = 1

5432102

15

1

05

02

15

1

05

aevsd

aevsu


25V level rise edge

nd

d

V(n)

FIFO queueTo

In

In 1

In

In

In

InIn

en

In

In

Out

Out

OutOut

Out

Out

Out In Out In

In

Out

Out

Out

Compare to 10

Cancel timeout




In1

In2

Signal scope



AE LEVELIn

InIn

Get attribute

Get attribute1

Level

Phase

Match

Chart

Entity sink

Signal scope1

Out

Out


+49

+16

minus18

minus51





00

200 400 600 800 1000 1200 1400 1600 1800

AE PHASE










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










5

10

15

20

25

30

Ope

ratin

g lif

etim

e (h

ours

)


S2D S2TS2E (5) S2E (10)S2E (20) S2E (25)

724784844

904964

1024

1084

1144

1204

1264

1324

1384

1444

1504

1564

1624

1684

1744

1804

1864

1924

1984

2044

2104

2164

2224

2284

2344

2404

2464

2524





0

5

10

15

20

25

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10)

S2E(20)

S2E(25)

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


Event sampling

Up-pulse (Rup)

s(t)


aei

Step-down (Rsd)Ω

S2D S2T S2E(5)

S2E(10)

S2E(20)

S2E(25)


02468101214

5

10

15

20

25

Ope

ratin

g lif

etim

e (h

ours

)


times103times102



Ope

ratin

g en

ergy

(120583J)

dur

ing1

s


S2D S2T S2E

Energy 445 418 315 238 185 160Lifetime 1714 4284 7947 9707 11456 12530

50100150200250300350400450500

Ope

ratin

g lif

etim

e (h

ours

)


S2E(5)

S2E(10) (20)

S2E(25)

0

2

4

6

8

10

12

14times103




Event samplingt

Rsd RsuRsu

aei


s(t) transmitted

sd(t) reflected

Ope

ratin

g en

ergy

(120583J)

dur

ing1

s




6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










6 Conclusion







Acknowledgments


References




























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation























RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Documents

Research Article Accuracy-Energy Configurable Sensor Processor …downloads.hindawi.com/journals/tswj/2014/546563.pdf · · 2015-11-23Accuracy-Energy Configurable Sensor Processor