Your Home as Doctor:

Smart Homes Enable Medical Diagnosis

D. Schreurs

KU Leuven, Belgium

Outline

• Motivation

• System Architecture

• Blocks’ Design & Experimental Results

• System Design & Experimental Results

• Durability

• Conclusions

2

Motivation

• Observation: ageing population

���� Importance of Ambient Assisted Living

= long-term health monitoring− Fall detection− Localization− Vital signs monitoring

3

0

1000

2000

3000

4000

5000

6000

wo

rld

po

pu

lati

on

(millio

ns)

2015 2050 2100

year

aged 60+

total

Source: bbc.com

Approach: Radar WSN

4

Base Station

Radar

Sensor

WRSN System Architecture

M. Mercuri, et al., "Analysis of an indoor biomedical radar-based system for health monitoring,"

IEEE Trans. Microw. Theory Techn., vol. 61, no. 5, pp. 2061-2068, May 2013. 5

Radar Technologies

typefrequency

bandwidth

ranging

meas.Doppler

complexity

(& costs)

CW narrowband NO YES low

UWB IR UWB YES NO high

SFCW/FMCW

UWB (& narrow band)

YES YES low-medium

f0+∆∆∆∆ff0 f0+(N-1)∆∆∆∆ff

CW SFCWCWCW UWB-IR

6

• fISM ���� fall detection

���� vital signs monitoring

• SFCW waveform ���� localization

f0 = 6 GHz

N = 40

∆∆∆∆f = 25 MHz

RMAX = 6 m

∆∆∆∆R = 15 cm

2.004 sec

T=100 µs2 sec

5.8 GHz

f0 f0+∆∆∆∆f f0+(N-1)∆∆∆∆f

5.8 GHz30 µs

Settling time

M. Mercuri, et al., “Optimised waveform design for radar sensor aimed at contactless health monitoring,”

Electron. Lett., vol. 48, no. 20, pp. 1255-1257, 27 Sep. 2012.

Radar Waveform

7

Outline

• Motivation

• System Architecture

• Blocks’ Design & Experimental Results

• System Design & Experimental Results

• Durability: Energy supply

• Conclusions

8

Radar WSN

9

Base Station

Radar

Sensor

Radar Sensor

M. Mercuri, et al., "Analysis of an indoor biomedical radar-based system for health monitoring,”

IEEE Trans. Microw. Theory Techn., vol. 61, no. 5, pp. 2061-2068, May 2013.10

Antenna Design

• In-door setting = sensor mounted to wall or ceiling

• Antenna design challenges

– avoid backscattering & crosstalk

– semi-spherical radiation pattern

– wideband requirement

TX

RX

target reflection

11

-50

-40

-30

-20

-10

0

1 2 3 4 5 6 7 8 9

Freq (GHz)

Sp

ara

m (

dB

)

S11

S21

Antenna Design

• Ground Plane: backscattering

• Metal Wall: crosstalk (S21)

12

Zigbee Radar

P. J. Soh, et al., “Dual-band Planar Bowtie Monopole for a Fall-Detection Radar and Telemetry System,”

IEEE Antennas Wireless Propag. Lett., vol. 11, pp. 1698-1701, Dec. 2012.

S. Yan, et al., “A low profile dual-band antenna loaded with artificial magnetic conductor for indoor radar systems,”

IET Radar, Sonar & Navigation, 7 pp., 2015.

Design based on Artificial Magnetic Conductor (AMC) layer

• smaller form factor

• no metal wall!

AMC Based Antenna Design

13

S. Yan, et al., “A low profile dual-band antenna loaded with artificial magnetic conductor for indoor radar systems,”

IET Radar, Sonar & Navigation, 7 pp., 2015.

AMC Based Antenna Design

AMC Antenna Reference Antenna

Size (mm3) 56×56×3.12 42×36×32

Low Freq. Band (GHz) 2.395-2.485 2.22-2.48

High Freq. Band (GHz) 5.78-8.89 5.8-6.77

Gain (dB) @ 2.45 GHz 5.05 5.7

Gain (dB) @ 6-7 GHz >7.0 >7.2

FBR (dB) & 2.45 GHz 15 0

FBR (dB) & 6-7 GHz >11 >7.5

Cross coupling (dB) <-25 <-25 (with wall)

FBR = Front-to-Back Ratio

14

Monostatic Radar Architecture

15

LO

RF

(target + unwanted)

baseband

VCOmismatch

isolation

15

Monostatic Radar Architecture

16

received

17

0 1 2 3 4 5 60

20

Range profile (m)

Ma

gn

itu

de

0 1 2 3 4 50

0.006

Range profile (m)

Ma

gn

itu

de

without leakage canceller with leakage canceller

target at 5 m

UNWANTED LEAKAGE TARGET DETECTED

Monostatic Radar Architecture

17

Base Station

18

Microcontroller-based DSP-based

C. Garripoli, et al., “Embedded DSP based Telehealth Radar System for Remote In-door Fall Detection,”

IEEE J. Biomed. Health Inform., vol. 19, no. 1, pp. 92-101, Jan. 2015.

• IFFT on SFCW baseband signals ���� absolute distances

f0 f0+Δf f0+2Δf … f0+(N-1)Δf

C1 C1 C2 … C1N

0 1 2 3 4 50

1

2

3

range profile (m)

ma

gn

itu

de

IFFT

Data Processing: Localisation

19

Data Processing: Localisation

• Compensation ���� remove ‘static’ reflections (e.g., furniture, crosstalk)

• Calibration ���� places the target in the right position

0 1 2 3 4 5 60

1

range profile (m)

mag

nit

ud

e

0 1 2 3 4 50

0,1

range profile (m)m

ag

nit

ud

e

target buried

target detected

Before compensation and calibration After compensation and calibration

crosstalk

Reducing backscattering and crosstalk is fundamental!

Example: Target at 4 m

M. Mercuri, et al., “A practical distance measurement improvement technique for a SFCW-based health monitoring radar,”

ARFTG Microwave Meas. Conf., Jun. 2013. 20

Data Processing: Localisation

Distance

Anechoic Chamber Real Room

% Person1 % Person2 % Metal % Person1 % Person2 % Metal

1 m 100 100 100 100 100 100

2 m 100 100 100 100 100 100

3 m 93.3 96.7 100 90 86.7 100

4 m 93.3 93.3 100 90 83.3 100

5 m 86.67 90 100 76.7 73.3 100

� 30 range profile measurements for each distance and for each target

21

P. Karsmakers, et al., “Automatic in-door fall detection based on microwave radar measurements,”

Eur. Radar Conf., 2012.

time frame

freq

uen

cy b

in

2 4 6 8 10 12 14

5

10

15

20

25

300.2

0.4

0.6

0.8

1

time framefr

eq

ue

ncy

bin

2 4 6 8 10 12 14

5

10

15

20

25

300.2

0.4

0.6

0.8

1

0 1 2-5

0

4

time (s)

sta

nd

ard

ize

d a

mp

litu

de

0 1 2-5

0

5

time (s)

sta

nd

ard

ize

d a

mp

litu

de

WALKING FALLING

Data Processing: Fall Detection

Classification approach: LS-SVM-GA

22

0 60 120

normal movement

falling

time (s)

falling

Real-Time Fall Detection

• Sliding window approach

= Segmentation

• Real-time operation

• ~316 ms maximum delay

C. Garripoli, et al., “Embedded DSP based Telehealth Radar System for Remote In-door Fall Detection,”

IEEE J. Biomed. Health Inform., 2014. 23

SPEED MEASUREMENTS:

− Sensor fixed to the wall or ceiling

− Falls over different orientations

− 3 subjects

Fall Detection: Sensor Positioning

24M. Mercuri, et al., “Analysis of a Fall Detection Radar placed on the Ceiling and Wall,” APMC, Nov. 2014. 24

1.7 2.2 2.70.8

1.6

2.4

time (s)

am

plitu

de (

V)

I channel

Q channel

2.5 3 3.51.1

1.6

2.1

time (s)

am

plitu

de (

V)

I channel

Q channel

3 3.5 40.6

1.6

2.6

time (s)

am

plitu

de (

V)

I channel

Q channel

1 1.5 21.4

1.6

1.8

time (s)

am

plitu

de (

V)

I channel

Q channel

2.8 3.3 3.81.4

1.6

1.8

time (s)

am

plitu

de

(V

)

I channel

Q channel

2 2.5 31.4

1.6

1.8

time (s)am

plitu

de (

V)

I channel

Q channel

falling

0°

falling

90°

walking

0°

WALL CEILINGWALL

� Falls have an important vertical motion component

� Doppler effect limitation

Fall Detection: Sensor Positioning

CEILING

25

Fall Detection: Sensor Positioning

Angle

( °°°°)

CEILING WALL

Subject 1 Subject 2 Subject 3 Subject 1 Subject 2 Subject 3

0 3 3 3 3 3 3

30 3 3 3 3 3 3

60 3 3 3 0 0 0

90 3 3 3 0 0 0

120 3 3 3 0 0 0

150 3 3 3 3 3 3

180 3 3 3 3 3 3

210 3 3 3 3 3 3

240 3 3 3 0 0 0

270 3 3 3 0 0 0

300 3 3 3 0 0 0

330 3 3 3 3 3 3

� Radar to the ceiling ���� to overcome Doppler effect limitation

− 3 mimicked frontal falls per target/position

26

Sensor Positioning: Breathing

27

Wall (Sensor 3 - lateral)Wall (Sensor 4 - frontal)

0 1 2 3 4 50

0.004

X: 0.3967

Y: 0.004436

Frequency (Hz)

Magnitude

0 1 2 3 4 50

0.02

X: 0.3662

Y: 0.02129

Frequency (Hz)

Magnitude

Person Sitting on a Chair

0 1 2 3 4 50

0.02

X: 0.3662

Y: 0.02051

Frequency (Hz)

Mag

nit

ud

e

0 1 2 3 4 50

0.02

X: 0.3357

Y: 0.02342

Frequency (Hz)

Mag

nit

ud

e

0 1 2 3 4 50

0.008X: 0.3357

Y: 0.00771

Frequency (Hz)

Mag

nit

ud

e

0 1 2 3 4 50

0.01

X: 0.3662

Y: 0.01426

Frequency (Hz)

Mag

nit

ud

e

→→→→ Clear benefit of having radar sensor network!

Prone

Supine

Ceiling (Sensor 1) Wall (Sensor 4)

Sensor Positioning: Breathing

28

Breathing Heart beat

Vital Signs Monitoring

29

Outline

• Motivation

• System Architecture

• Blocks’ Design & Experimental Results

• System Design & Experimental Results

• Durability

• Conclusions

30

Radar WSN

31

Base Station

Radar

Sensor

Wireless Radar Sensor Network

• Challenges: synchronization & avoiding interference

� Radar operation

�Speed: FDM among radar sensors

�Localisation: TDM

→ one sensor at a time

� Wireless communications

�TDM: Zigbee and radar sensing

�TDM: 4 ms delay among sensors

���� THE SENSORS VIRTUALLY OPERATE AT THE SAME TIME!!!

M. Mercuri, et al., “Dual-Mode Wireless Sensor Network for Real-time Contactless In-door Health Monitoring,”

IMS, 2015. 32

Experimental Setup

33

���� Errors within the radar resolution of 15 cm!

Localisation

34

Experimental Setup

35

Fall DetectionSensor 1

Sensor 2

Sensor 3

Sensor 4

• Fall is not detected by

Sensor 2

• Multiple sensors expand

coverage range and

improve detection

reliability!

36

Outline

• Motivation

• System Architecture

• Blocks’ Design & Experimental Results

• System Design & Experimental Results

• Durability: Power supply

• Conclusions

37

Smart Homes: Durability

38

Sensor Power consumption

Zigbee node 40 mW

Temperature sensor 0.5 mW

IR motion detector 0.3 mW

Biomedical sensor > 100 mW

Power needs

(order of magnitude)

Power consumption radar sensor

• using off-the-shelf components:

– wireless communications (Zigbee): 40 mW

– radar sensing: >100 mW

– microcontroller: 100 mW

• using dedicated chip design: 50 mW - 100 mW expected

39

Smart Homes: Energy Needs

Smart Homes: WPT?

40

≠ RFID

Base station Sensor Node

Wireless Power

Control Data

Application Data

Challenge: Single antenna @ Sensor Node

Regulations

Frequency Bands for Non-Specific Short Range Devices in Europe

Ref: Electronic Communications Committee ����

Single base station is not feasible

41

Base stations

Sensor Node

Wireless Power

MISO WPT

42

MIMO WPT

Base stations Sensor Nodes

Wireless Power

43

WPT & Data Receiver

X. Zhou, et al.,

IEEE Trans. Comm.,

2013.

T.J. Lee, masterthesis,

KU Leuven, 2016.

Boost regulator

ADC

44

Experimental Results

• Modulation scheme such that received symbols are clear

• Received data signal is based on baseband intermodulation

products of multisine RF excitation

T.J. Lee, masterthesis,

KU Leuven, 2016.

45

Conclusions

Wireless radar sensor networks:

= viable technology for Smart Homes!

46