1992-8645 A NOVEL PHOTOPLETHYSMOGRAPH SENSOR …examinations. Furthermore, simple software computation and low power consumption. Practical application of this study will be presented

Journal of Theoretical and Applied Information Technology15th December 2016. Vol.94. No.1

© 2005 - 2016 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

142

A NOVEL PHOTOPLETHYSMOGRAPH SENSOR PROBEDESIGN FOR MOTION ARTIFACT DETECTION: A

COMPARISON STUDY WITH THREE AXISACCELEROMETER

1 MUHIDEEN ABBAS HASAN, 2 FAHMI SAMSURI, 3 KOK BENG GAN1 Faculty of Electrical and Electronics Engineering, Universiti Malaysia Pahang, 26600, Pahang, Malaysia

2 Department of Electronics, Technical Institute/Dour, Northern Technical University, Foundation of

Technical Education, Iraq3 Department of Electrical, Electronic, and Systems Engineering, Faculty of Engineering and Built

Environment, Universiti Kebangsaan Malaysia, Bangi, Malaysia.

E-mail: [email protected], [email protected], [email protected]

ABSTRACT

Photoplethysmography (PPG) sensors have become widespread in most of the healthcare categories; itsdrawback is unreliable during non-stationary states. Adaptive noise cancellation is one of the severaltechniques have existed to address this issue. The problem of implementing this method still lacks for totaldetection of induced noise due to the motion. In this work, a new method was introduced for noise detectionbased on novel PPG sensor probe design by adding a covered photodetector (CPD) as a motion artifactreflector, in addition to the main photodetector (MPD). Experiments of several motion kinds were executedto the tied (CPD) with the accelerometer. Data analysis for the collected signals showed a lengthyconvergence between the amplitude fluctuations in the time domain. The CPD precisely indicated all thefundamental frequencies of various induced noise, unlike the accelerometer. Using these photodetectors,the CPD as noise reflector and the MPD to track the contaminated PPG signal is more reliability than otherapproaches. Implementing this method ensures a high-level accuracy and reducing the cost of repeated falseexaminations. Furthermore, simple software computation and low power consumption. Practical applicationof this study will be presented in our next work.

Keywords: Photoplethysmograph, Motion Artifacts, Accelerometer, CPD, MPD

1. INTRODUCTION

Photoplethysmography is a noninvasive methodto measure changes in arterial blood flow throughthe skin utilizing infrared light and photodetector.PPG technique has been accepted by theInternational Standards Organization (ISO) and theEuropean Committee for Standardization as thestandard non-invasive measure of oxygen saturationlevel since 1987[1]. PPG technique has beenimplemented in many different medical settings,including clinical, physiological monitoring (bloodoxygen saturation level, Heart Rate, BloodPressure, Cardiac Output, Respiratory Rate), FetalHeart Rate and person’s identity [2-5]. Despite themany PPG signal advantages as an easy to set up,comfortable, inexpensive compared to the othertypes of such medical devices [6]. However, PPG

signals could be easily affected by intended orunintended motion in the measurement processwhich may conducive to the incorrect interpretationof the reported PPG signal waveform which can beused for diseases classification [7, 8]. So it is notstraightforward to remove the motion artifacts fromthe PPG signal.

Valuable efforts were spent to cope with resultedmistakes in measurements during motion state byusing different techniques as reported by [9-13]. Ina comparative study [14] of several currents usedmethods for the PPG signal processing while thenon-stationary state was concluded that each ofthose methods has its advantages anddisadvantages. Also, the mentioned study indicatedthat the ANC technique is not the worst of theothers and could be better if a proper hardware




143

provided the reference noise signal instead of usingthe synthetic way, as the both are traditionallyutilized for this task.

What concerns us is the ANC method; the maindrawback of its implementation is providing theactual reference noise signal as a major factor in thequality of the adaptive filter performance. Usually,it is provided by two approaches an additivehardware or synthetic noise.

Also, the Independent Component Analysis(ICA) technique [10] was used to address the sameissue depending on statistical independenceassumption between motion artifact and heartpulsation. The difficulty of the ICA application liesin the necessity of providing multiple pre-processedsignals to overcome the interrelationship betweenheartbeat rhythm and the motion artifact [15].However the preferred application is for ANC beingbased on fewer complexity calculations, the bothtechniques may do the same effective performancein case of producing entirely separate motionartifact.

An added another transducer such 3-axisaccelerometer attached in the same area of thesubject test to reflect the effect of motion. Manyresearchers concluded that there is no significantrelationship between motion artifact and accelerateddata from the added accelerometer in PPG sensorand what associated with extra PPG sensor,additional sources of noise can be caused [16-19].

Another Approach is by synthetic noise throughthe use of PPG signals itself that recorded fromwearable oximetry devices to mathematicallyestimate the reference noise as being employed instudies [20, 21]. The synthetic noise approach inmentioned studies is not efficient to meet robust andsudden motions during performing some physicalexercises, furthermore it is computationallyintensive, which may not be suitable for systemswith limited resources [22, 23].

Hence, the need for reliable approach ensures theentire detection for the induced noise due to anymotion, is still in need for safe implementing of thePPG sensor leads to a precise monitoring of variousheart diseases during daily motion life.

In this paper, we demonstrate a new approach toproviding the noise reference signal highlycorrelated to the induced motion effects PPG signaldue to the motion artifact, based on modifiedtraditional PPG sensor. The idea was built on priorknowledge of adaptive filter [24] and PhotodiodeCharacteristics [25]. The modification for classic

PPG sensor is introduced by adding a coveredphotodetector CPD away from light effect as anoise reflector (noise reference signal in ANC, inaddition to the main photo detector (MPD) which iscustomized to provide the corrupted PPG signal(primary signal in ANC). This approach ensurescompletely same source for both signals and avoidsthe electronic noise effect, which usually isneglected on traditional methods for such sensitivesignal processing. Validation of the fact that theCPD as a reflector for motion artifact would be in amanner of fastening together the accelerometer andthe CPD, and exposed to different movements;along with X, Y and Z axis, also during silencestate. Calculating the resulted two signals of CPDand the combined signal of 3-axis accelerometersignals for all mentioned activities, and thenanalyzing amplitude fluctuate in the time domainand Fast Fourier Transform FFT peak plots in thefrequency.

2. METHODOLOGY

2.1 MotivationThe conflict between researchers to address the

fundamental obstacle of unreliable measurementsduring motion state for the PPG sensor motivatedus to investigate the real reasons stand behind thisdrawback in more realistic perspective. Some of theresearchers adopted the added hardware like theaccelerometer and others went to utilize differentalgorithms to compensate the motion effect inprocessing the corrupted PPG signal withoutachieving the optimum accurate. Therefore weinvolved a new concept of similar twin-photodetectors as two eyes, one to generate thedistorted PPG signal while the other one would becovered up from the light effect to reflect the samedistortion accompanied by the distorted signalsimultaneously. As an efficient solution to providea proper environment for ANC and ICAApplications, and to avoid the electronic noiseemergence in the case of adding extra hardware,also the occurred frequency overlap that preventsthe exact algorithmic estimations.

2.2 Background of StudySince the photodiode is the critical component in

generating the PPG signal, also it exposes tovarious types of noise, the manufactured PPGsensor is an electronic instrument subjects toEquipartition theorem [26]. This theorem statesthat: (Each degree of freedom of any systemfluctuates with energy equal to 1/2 kBT=2*10-21 J)at the standard temperature of 290 K. (where kB=1.38 x 10-23 J/K, is the Boltzmann constant and T




144

the absolute temperature). Motions give rise to arandom electric current even in the absence of anexternal electrical power source. Electronic noise inelectronic systems constraints the wholeperformance of the systems. Therefore, there is aparamount correlation between the motion artifactand the induced electronic noise in photodetectors.

From the photodiode voltage-currentcharacteristic, when a reverse bias is connected areversed saturation current I sat does exist, it relatesto the generated dark current Id as being calculatedin equation (1) and highlighted in figure 1 [27-29].

Where Id is the photodiode dark current, Isat isthe reverse saturation current relates to the darkcurrent; q is the electron charge. VA is the appliedbias voltage, KB = 1.38 x 10-23 J/K, is theBoltzmann constant, and T is the absolutetemperature (273 k = 0 °C).

Figure 1. Characteristic I-V Curves Of AnOptoelectronics Photodiode For Reverse And ForwardBias Modes Of Operation. P0-P3 Represents DifferentLight Levels, Starting From Dark State And Up.

(2)

Hence during reverse bias and no incident light, asmall dark current occurs and with increasing theincident light, the generated photocurrent IPh willalso increase depending on incident light power Pas in equation (2). Also from the photodiode noisecharacteristic, the photodiode noise acts the inducednoise because of motions effect that raises therandom motion of electron holes to flow due toincident light and dark current effect. Thephotocurrent noise can be specified in two mainsources; Shot Noise (ISN) which exists as a resultof the statistical fluctuations during the generated

photocurrent in both forward and reverse bias inaddition to the Johnson noise (IJN) which is relatedto shunt resistance available in electronicphotodetector circuit. The total induced noise (ITN)in photodetectors can be calculated by the (equation3) [25].

(3)

2.3 Novel PPG Sensor Probe DesignExploiting the similar property of photodetector

in the presence of light or dark state [25] led toreconsider the classical PPG sensor probe design.Figure 2 shows the block diagram of the new PPGsensor probe design with two reverse biasedphotodetectors in the same position setting wouldbe a meaningful solution for PPG signal processingin ANC technique.

Figure 2. Block diagram of the new PPG sensor probedesign.




145

Figure 3. (a) is the required PCB for PPG sensor probe,(b) is the PCB design for suggested sensor probe, (c) and

(d), are the new and classical PPG sensor layoutrespectively.

One of the photodetectors will be covered upfrom light effect, and light will illuminate only thesecond one. The occurred photocurrents in bothphotodetectors meet the same induced noise, but thedifferent is the bright one will contain the PPG dataas well. In such a way we would provide corruptedPPG data by the illuminated photodetector and theinduced noise by the covered one. Utilizingadaptive noise cancellation with the minimumrequired computation algorithm free-noise issufficient, and the PPG signal would be obtainedaway from the complexity of estimation process aswell.

The idea of presenting this design is to producean alternative method for routine use of 3-axisaccelerometer as a reference sensor to provide thenoise reference signal in applying the ANCtechnique. In the design part, a traditional PPGsensor was modified to be with one infra-red LEDand two amplified photodetectors (APDs) type(OPT101) having been connected in parallel as inFigure 1. The generated photocurrent IM which isproportional to the light power and wavelength ofincident infrared light would be converted to avoltage by the configured trans-impedance of 1MΩ.The resulted voltage forms the output VM of themain amplified photodetector A (MAPD - A) whichgives the corrupted PPG signal and the motionartifact effect in addition to the electronic noise.

The another output VC of the covered amplifiedphotodetector B (CAPD - B) which results bypassing the generated dark current IC through the

trans-impedance of 1MΩ, would give the inducedmotion artifact noise that has been assumed, theyare the same portion in both amplifiedphotodetectors. Both photodetectors are in reversedbiased (photoconductive mode). VC is completelyunlike the VM on a paramount issue that they arenot correlated in term of PPG while in the durationof the motion artifact noise tends to be extremelycorrespondent. The aim of this design is to obtainthe reference signals for adaptive noise cancellationprocessing later. Typically, the adaptive filter is fedwith two types of signal. Primary signal is thedesired signal contaminated by noise while thesecond represents the reference signal of noise thatcorrupted the wanted signal. The primary signal canbe characterized as in equation (4) due to resultedoutput from the main amplified photodetector(MAPD).

VM from MAPD (A) = Clean PPG signal + MotionArtefact + Electronic Noise (4)

Whereas the noise reference signal can becharacterized due to the output voltage from thecovered amplified photodetector (CAPD) as inequation (5).

VC from CAPD (B) = Motion Artefact + ElectronicNoise (5)

With these characterized signals adaptive filter cando the task of filtering contaminated signal andupdate its coefficient upon the provided noisereference signal.

3. VALIDATION STUDY

To justify the use of the cpd to determine the realeffect of motion artifact during implementing theANC for PPG signal processing, instead of 3-axisaccelerometer which is customized to indicate onlythe motion according to gravity influence. Twotypes of analyzing have been done. Series ofexperiments were conducted using both sensors andwere exposed to different kinds of movement alongwith x, y, and z axis. The resulted data from twosources, the CPD, and the acceleration wererecorded for further analyzing as explained in (3.3).

3.1 Data Acquisition Set-up and ProtocolData of 3-axes accelerometer (AD335, Analog

Devices) were collected separately for each X, Y,and Z axis in addition to the data from a coveredphotodetector, utilizing the four channels DAQ-NI9215 as shown in Figure 4b.




146

Figure 4. (a) is the state of the main photo detector(MPD) and covered photodetector during motion

execution, (b) is the setup of data acquisition.

All data were sampled at 100 Hz and 300 samplesper channel using LabVIEW 2015 (NationalInstruments Corporation). The three channels of X,Y, and Z axis have been combined into r due to theequation (6) into one signal.

(6)where r, is the combined signal of the 3-axisaccelerometer separated channel signals x, y and z.All the utilized hardware and software parametersof resulted in this approach layout, as revealed insections of design and validation, are summarizedin Table 1, as following.

Table 1. Key intrinsic for implemented hardware andsoftware parameters.

I. Key DesignParameters

II. Types III. Values

IV. TraditionalPPG sensor

V. NellcorDS100A

VI. –

VII. Amplifiedphotodetector

VIII. OPT101 IX. –

X. 3-axesAccelerometer

XI. AD335 XII. –

XIII. Four channelsDAQ.

XIV. NI 9215 XV. –

XVI. Sampling rate XVII. – XVIII. 100Hz

XIX. Samples perChannel

XX. – XXI. 300samples

XXII. Butterworthlow-pass filter

XXIII. XXIV. 15Hz

Data were collected in two states during motion andsilence state. Three motion types by moving thehand holding the both devices along with X, Y, andZ axes were executed as in figure 4a. For the

second state, devices were left on the table withoutany effect of a fan or air-condition in an attempt tokeep them under the same external impacts as infigure 4b. After that, the acquired data was filteredby a second order Butterworth low pass filter of 15Hz. Then, the signals were plotted to observe thematch of the recorded signals during the amplitudefluctuate in the term of time domain for the motionand no motion states. Finally, both cases weretransformed into frequency domain using FFT forfrequency domain analysis.

3.2 Amplitude FluctuationsThe fluctuations of all signals amplitude that

involve the motion of X, Y, Z, r and CPD signalwere monitored during three types of relatively fastmotion of frequent around 6 Hz in term of time.The kind of trends of these fluctuations can indicatethe primary relationship among them.

3.3 Fast Fourier TransformThe Fast Fourier Transform (FFT) is a

mathematical way to convert a function of a timedomain into frequency domain. Fourier analysis(FA) of a periodic function indicates the extractionof series sine and cosine waves to reconstruct thefunction [30, 31]. The equation (7) defines theFourier Transform:

(7)

where x(t) is the time function, x(f) is the fastFourier transform function and (ft.) is the analyzingfrequency. Calculating the FFT peaks of the threechannels of 3-axis ACC and their combined signalin addition to the covered photodetector signal, thencomparing the resulted peaks must be matched toeach other in same frequency domain.

4. RESULT AND DISCUSSIONS

4.1 Time Domain AnalysisDuring motion state, 3-axis acceleration. X, Y, Z,

resultant acceleration r and the CPD, theiramplitude trends for continuous 3 seconds datawere observed. Figure 5 of 3 windows showed theamplitude fluctuations during the motion along theX axis as an example, where the data was filtered,and the DC component removed. Despite, the smalldifference of other intensive fluctuations in CPDsignal, belongs to the variance noise property ofeach device within time [27]. Although, theamplitude fluctuations indicate a noticedconvergence between the two approaches asreflectors for the motion effect. From the bottom




147

part of the mentioned figue, it is clearly noticed thatthe magnitude of accelerometer amplitude washigher than CPD amplitude as it is designed formeasuring motion and follows the Gravity impactduring movement within the three axes.

For the silence state, the both devices outputsfluctuate as a result of the environment effect suchelectromagnetic devices which also results in anelectronic noise [28, 29] differs from device toanother as shown in figure 6. In this state, whatdrawn our attention is that the amplitude magnitudeof CPD is unlike in the previous state of motionbecause the dark current Id effect always exists. Forsuch instant, the generated dark current

Figure 5. Top And Middle Windows For Raw Data OfAcceleration And CPD Signals, The Bottom, Is The

Filtered Signals With Removed D.C Component.

Id would represent the whole induced noisewhich is the same portion simultaneously will beinduced in MPD of the proposed design.

Figure 6. Amplitude fluctuation of both acceleration andCPD signals in silence state for continuous 3 seconds.

4.2 Frequency Domain Analysis

In addition to what we had discussed about theamplitude fluctuate in the two different states,correspondence of plotted FFT peaks for the CPDand 3-axis acceleration in the silence, state showshow the both devices approximately share the sameharmonic spectrum amplitudes. Although in amotion state they also share the same plottedspectrum in the frequency domain, they aredifferent in term of the energy signals distribution.Figure 7 Highlights the close convergent of totalfundamental harmonics as it is pointed by arrowsfor maximum and minimum in 20 Hz duringsilence state for continuous 3 seconds.

Figure 7. FFT Plot During Silence State For BothCovered Photodiode And Accelerometer.

In the (Figures 8, 9 and 10) the FFT peaks wereplotted for acceleration and CPD signals in thesame frequency band and time for the previousstate for motions along with X, Y, and Zrespectively. For the first motion plot along with Xaxis figure 8 (top) indicates that acceleration devicehas the precedence task of determining thespectrum of the movement as a customized sensorfor this limited function. Whereas in the bottompart of the same figure, CPD FFT peak plot haspointed all the affected frequencies according to theeffect priority of various impacts and not only themotion effect, unlike for the acceleration plots.




148

Figure 8. FFT plot during motion along with X-axis forboth covered photodiode and accelerometer.

The same concept can be deduced by examiningthe (Figures 9 and 10). Which belong to motionsalong with Y and Z axes, even more than thespecifical accuracy for each harmonic that canexceed the expected frequencies overlap as revealedin figure 9 bottom. So undoubtedly could be guidedby the existed results that the best solution todetermine the correspondence noise in ANCapplications is the recruitment of the same typesource of active ingredient in measurementsrequired high accuracy.

In previous study [32] two photodetectors wereused to indicate two measurements at the same timefurthermore the novel work in [33] utilized twoLEDs of different wavelength lights (Red andGreen) and one photodetector to get two signalvalues according to switching time control. Butboth studies have adopted the comparativedifferential processing between the collectedsignals to remove the occurred noise. The reportedresults were limited to the motion speed due to theneed of constant equal amplitudes for generatedsignals as precondition for an efficient applicationof the differential processing approach to removethe noise and this is not easy be achieved duringmotion.

Figure 9. FFT plot during motion along with Y-axis forboth covered photodiode and accelerometer.

Figure 10. FFT plot during motion along with Z-axis forboth covered photodiode and accelerometer.

Application of this approach results in anessential quality of accuracy; that can assist inminimizing the complex processing to determiningthe desired signal. Also allow for high reliableoutcomes by reconstruction some of the earlierstudies for instance in a sophisticated study [34], tocalculate hemoglobin content in blood, where FFTmethod was used to remove the redundant parts ofinaccurate PPG signal which already has its cons.

5. CONCLUSION

The achieved findings of amplitude fluctuationsand FFT peak plots showed that the proposedapproach is capable of indicating all kinds ofeffective noise caused by motion due to thesimilarity of influential sources for the PPG signalprocessing within ANC technique. Also provedthat the traditional methods such 3-axes




149

acceleration or an extra PPG sensor are not quiteeffective to provide the real induced noise becauseof the different source type for the first case and thedifficulty of uniting screen room of skin subject, forthe second instance. The outcomes of utilizing suchmethod are to ensure the high accuracy and reliablemeasurements during diagnosis of pathologicalsituations and reduce false alarms. In addition toreducing the expenses of the repeated examinationswhen the lack of reliability and economic exchangeof energy compared to the aforementioned usedmethods where the first consists of three channelsand the second includes a LED and detector.Practical application of this approach besidesanother comparison study among the linearcorrelations for the collected signals from CPD,MPD, and the acceleration as further evidence willbe in our future work.

REFRENCES:

[1] A.B. Shang, R.T. Kozikowski, A.W.Winslow, and S. Weininger, Developmentof a standardized method for motiontesting in pulse oximeters. Anesthesia &Analgesia, 2007. 105(6): p. S66-S77.

[2] K.B. Gan, E. Zahedi, and M.A.M. Ali,Transabdominal fetal heart rate detectionusing NIR photopleythysmography:instrumentation and clinical results. IEEETransactions on Biomedical Engineering,2009. 56(8): p. 2075-2082.

[3] V. Jayasree, Selected cardiovascularstudies based on photoplethysmographytechnique. 2009, Citeseer.

[4.] S.C. Millasseau, J.M. Ritter, K. Takazawa,and P.J. Chowienczyk, Contour analysis ofthe photoplethysmographic pulsemeasured at the finger. Journal ofhypertension, 2006. 24(8): p. 1449-1456.

[5.] S.N.A.M. Azam and K.A. Sidek, TimeVariability Analysis ofPhotoplethysmogram BiometricIdentification System. Indian Journal ofScience and Technology, 2016. 9(28).

[6] P.Y. Cheang and P.R. Smith, An overviewof non-contact photoplethysmography.Dept. of Electronics & ElectricalEngineering, Loughborough University,LE, 2003. 1(1).

[7] N.I. Zainal and K.A. Sidek, Real TimePPG Data Acquisition with GUI basedApplication for HRV Measurement. IndianJournal of Science and Technology, 2016.9(28).

[8] P.V. Prabhu, J. Sivaraman, S. Sathish, S.Vinurajkumar, and K. Manikandan,Detection and Evaluation of VascularWall Elasticity usingPhotoplethysmography Signals in SinusRhythm Subjects. Indian Journal ofScience and Technology, 2016. 9(2).

[9] M.R. Ram, K.V. Madhav, E.H. Krishna,K.N. Reddy, and K.A. Reddy. On theperformance of Time Varying Step-sizeLeast Mean Squares (TVS-LMS) adaptivefilter for MA reduction from PPG signals.in Communications and Signal Processing(ICCSP), 2011 International Conferenceon. 2011. IEEE.

[10] B.S. Kim and S.K. Yoo, Motion artifactreduction in photoplethysmography usingindependent component analysis. IEEEtransactions on biomedical engineering,2006. 53(3): p. 566-568.

[11] C. Lee and Y. Zhang. Reduction of motionartifacts from photoplethysmographicrecordings using a wavelet denoisingapproach. in Biomedical Engineering,2003. IEEE EMBS Asian-PacificConference on. 2003. IEEE.

[12] K.A. Reddy, B. George, and V.J. Kumar,Use of fourier series analysis for motionartifact reduction and data compression ofphotoplethysmographic signals. IEEETransactions on Instrumentation andMeasurement, 2009. 58(5): p. 1706-1711.

[13] K.A. Reddy, Novel methods forperformance enhancement of pulseoximeters. 2008, Ph. D. dissertation, Dept.Elect. Eng., IIT Madras, Chennai, India.

[14] K.V.P. Naraharisetti, M. Bawa, and M.Tahernezhadi. Comparison of differentsignal processing methods for reducingartifacts from photoplethysmographsignal. in Electro/Information Technology(EIT), 2011 IEEE InternationalConference on. 2011. IEEE.

[15] M. R. Ram, K. V. Madhav, E. H.Krishna, N. R. Komalla, K. Sivani, and K.A. Reddy, "ICA-based improved DTCWTtechnique for MA reduction in PPGsignals with restored respiratoryinformation," Instrumentation andMeasurement, IEEE Transactions on, vol.62, pp. 2639-2651, 2013




150

[16] L.B. Wood and H.H. Asada. Low varianceadaptive filter for cancelling motionartifact in wearable photoplethysmogramsensor signals. in 2007 29th AnnualInternational Conference of the IEEEEngineering in Medicine and BiologySociety. 2007. IEEE.

[17] S.H. Kim, D.W. Ryoo, and C. Bae.Adaptive noise cancellation usingaccelerometers for the PPG signal fromforehead. in 2007 29th AnnualInternational Conference of the IEEEEngineering in Medicine and BiologySociety. 2007. IEEE.

[18] H. Han, M.-J. Kim, and J. Kim.Development of real-time motion artifactreduction algorithm for a wearablephotoplethysmography. in 2007 29thAnnual International Conference of theIEEE Engineering in Medicine andBiology Society. 2007. IEEE.

[19] H.H. Asada, P. Shaltis, A. Reisner, S.Rhee, and R.C. Hutchinson, Mobilemonitoring with wearablephotoplethysmographic biosensors. IEEEEngineering in Medicine and BiologyMagazine, 2003. 22(3): p. 28-40.

[20] Z. Zhang, Photoplethysmography-basedheart rate monitoring in physical activitiesvia joint sparse spectrum reconstruction.IEEE Transactions on BiomedicalEngineering, 2015. 62(8): p. 1902-1910.

[21] R. Yousefi, M. Nourani, S. Ostadabbas,and I. Panahi, A motion-tolerant adaptivealgorithm for wearablephotoplethysmographic biosensors. IEEEjournal of biomedical and healthinformatics, 2014. 18(2): p. 670-681.

[22] A. Temko. Estimation of heart rate fromphotoplethysmography during physicalexercise using Wiener filtering and thephase vocoder. in 2015 37th AnnualInternational Conference of the IEEEEngineering in Medicine and BiologySociety (EMBC). 2015. IEEE.

[23] C. Zhou, J. Feng, J. Hu, and X. Ye, Studyof Artifact-Resistive Technology Based ona Novel Dual PhotoplethysmographyMethod for Wearable Pulse RateMonitors. Journal of medical systems,2016. 40(3): p. 1-10.

[24] K.B. Gan, E. Zahedi, and M.A.M. Ali,Application of Adaptive NoiseCancellation in Transabdominal FetalHeart Rate Detection UsingPhotoplethysmography. 2011: INTECHOpen Access Publisher.

[25] O. Optoelectronics, PhotodiodeCharacteristics and Applications. 2013,Last checked.

[26] J. Repond, J. Yu, C. Hawkes, Y. Mikami,O. Miller, N. Watson, J. Wilson, G.Mavromanolakis, M. Thomson, and D.Ward, Design and electronicscommissioning of the physics prototype ofa Si-W electromagnetic calorimeter for theInternational Linear Collider. Journal ofInstrumentation, 2008. 3(08): p. P08001.

[27] F. Bonani and G. Ghione, Noise insemiconductor devices, in Noise inSemiconductor Devices. 2001, Springer. p.1-38.

[28] J. Choma, Electronic noisecharacterization− Part I: System conceptsand theory. University of SouthernCalifornia, Los Angeles, CA, TechnicalReport# USC, 2011: p. 02-511.

[29] J.-M. Liu, Photonic devices. 2009:Cambridge University Press.

[30] C.P.L. Inspection, Increase MeasurementAccuracy by 10 Times with SignalConditioning. 2002.

[31] A.J. Wilder, Modal Analysis with theMobile Modal Testing Unit. 2013.

[32] Kuboyama Y, Motion artifact cancellationfor wearable photoplethysmographicsensor: Massachusetts Institute ofTechnology; 2010.

[33] Zhou C, Feng J, Hu J, Ye X. Study ofArtifact-Resistive Technology Based on aNovel Dual PhotoplethysmographyMethod for Wearable Pulse RateMonitors. Journal of medical systems.2016;40(3):1-10.

[34] A.M. Abbas, S. Ashok, S.P. Kumar, and P.Balavenkateswarlu, HaemoglobinDetection in Blood by Signal to ImageScanning using Photo-Plethysmo-Graphic-Technique (PPG). Indian Journal ofScience and Technology, 2016. 9(1).

Documents

1992-8645 A NOVEL PHOTOPLETHYSMOGRAPH SENSOR …examinations. Furthermore, simple software computation and low power consumption. Practical application of this study will be presented