Model Design and System Implementation for the Study of

Anti-motion Artifacts Detection in Pulse Wave Monitoring

Cong-Cong Zhou

, Jing-Yi Wang

, Li-Ping Qin

and Xue-Song Ye

1,2

College of Biomedical Engineering and Instrument Science, Biosensor National Special Laboratory,

Zhejiang University, Hangzhou 310027, P.R. China

State Key Lab of CAD&CG, Zhejiang University, Hangzhou 310027, P.R. China

Zhejiang Institute of Medical Device Testing, Hangzhou, China

Keywords: Photoplethysmography, Motion Artifact, Health Monitor, Pulse Wave Monitoring Platform (PWMP).

Abstract: Photoplethysmography (PPG) is a widely used technology for health monitor based on pulse wave

measurement by monitoring the blood volume of blood vessels via electro-optic technique. As a kind of non-

electrophysiological signal with low amplitude and low frequency, PPG signal may be easily disturbed by

motion artifact. This paper proposes a simulation method based on a new reflection model which includes a

skin-friendly flexible substrate with a narrow-band full-reflection film plating on it and an embedded system

accordingly to study anti-motion artifacts detection in pulse wave monitoring. Monte-Carlo method is

presented to simulate the dynamic human skin model and the results demonstrate the effectiveness of the

proposed model. A wrist worn artifact-resistive pulse wave monitoring platform (PWMP) is presented

accordingly, the measurement accuracy of pulse rate by the platform is within ±2 beats per minute(bpm) at

the range of 30bpm to 240bpm compared with the output of Fluke Index2 (produced by Fluke Corp, USA) in

stationary situation. Three kinds of typical postures are performed to verify the proposed model

experimentally, results show that the proposed platform has good correlation as compared to PC-60B Medical

Pulse Oximeter from Heal Force in the measurement of pulse rate, and the pearson correlation coefficient is

0.953(p<0.01), which reveals that the proposed model has the potential to recover pulse wave signal for pulse

rate monitoring.

1 INTRODUCTION

It is of great significance for the monitoring of pulse

rate, blood oxygen saturation, blood pressure (Zhou

et al.,2015). As one of the vital physiological

parameters of human body, pulse can reflect many

physiological and pathological features of the human

respiratory system, cardiovascular system, etc

(Nogami et al.,2018). Photoplethysmography (PPG)

is a method that measures the blood volume of blood

vessels via electro-optic technique. Accurate

acquisition and processing of PPG signal play an

important role in medical diagnosis, exercises and

other fields(Davoudi et al.,2014). As this method is

non-invasive, safe, reliable and adaptable, it is widely

used in physiological signal monitoring equipment,

especially wearable devices (Ra,2016; Zhou,2014),

for the monitoring of pulse rate and blood oxygen.

Currently, there are many devices designed to

detect the pulse rate, but not many of those have the

ability to detect pulse rate precisely. PPG signal,

which can be affected easily by motion artifact, is a

kind of low- frequency and weak non-

electrophysiological signal. Devices can perform well

in static conditions, but the precision is challenged

during movement due to the frequency overlapping

between pulse wave and motion artifact. Motion

artifact is mainly introduced in two kinds: one is the

interference that human body’s DC components

(muscles, tissue fluid, capillaries, etc.) introduce

during pulsation, such as capillary filling and tissue

fluid increase caused by increased metabolism during

exercise, etc. The other is the air gap generated by the

relative displacement between the optical sensor and

human body during exercise, which can be

considered as additive noise, many works aimed to

reduce this part as to enhance the anti-motion artifacts

ability of the detection devices. For example,

Fukushima et al., (2012) used an acceleration sensor

to obtain the reference signal of motion artifact. Zhou

et al.(2016) applied a differential channel following

green and red light PPG channels to enhance the anti-

102

Zhou, C., Wang, J., Qin, L. and Ye, X.

Model Design and System Implementation for the Study of Anti-motion Artifacts Detection in Pulse Wave Monitoring.

DOI: 10.5220/0008943201020109

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 1: BIODEVICES, pages 102-109

ISBN: 978-989-758-398-8; ISSN: 2184-4305

motion artifacts ability of the device. Shimazaki et al.

(2014) used another PPG sensor without contacting

the skin to detect the reference signal. Their works

showed that progresses had been made to the

cancelation of motion artifact. And to the best of our

knowledge, there is no similar research yet that

combines the model design(simulation) and system

implementation accordingly for anti-motion artifacts

detection in pulse wave monitoring based on PPG, the

systems mentioned usually based on priori

knowledge or through a lot of experiments and

measurements(Shimazaki et al.,2014 ).

This paper proposes a simulation method base on

a skin-friendly flexible substrate with a narrow-band

full-reflection film plating on it and an embedded

system accordingly to study anti-motion artifacts

detection in pulse wave monitoring based on PPG.

We present a new reflection type sensor model by

designing a narrow-band full-reflection film plating

on a skin-friendly flexible substrate to extract the

motion artifact separately, so as to achieve the

relatively accurate reference signal for further signal

processing. Meanwhile, this paper proposes a method

for modeling PPG signal based on MC (Monte-Carlo)

approach to obtain the simulation waveform from the

model mentioned above. To verify the effectiveness

of the proposed model, a wrist worn artifact-resistive

pulse wave monitoring system (PWMP) is presented

accordingly, three kinds of typical postures are

performed and the results show that the proposed

platform has good correlation as compared to the PC-

60B in the measurement of pulse rate.

2 METHOD AND MATERIAL

2.1 A New Reflection Type

Anti-motion Artifacts Model

In reflection type PPG detection, the Light Emitting

Diode (LED) and the Photo-Diode (PD) are located

on the same side of the tested part, and the change of

the light absorption (PPG signal) is obtained by

detecting the signal from PD. In the case of stationary

state, it is generally considered that the probe

composed of LED and PD is closely attached to the

surface of the human skin, and the method can obtain

accurate results. While in the case of motion, there

will be a relative displacement between the probe and

the measured part of the human body, which

introduces a layer of air whose thickness changes

with time. The air layer also absorbs and scatters the

light emitted from LED, thus motion artifact is

introduced into PD.

In this paper, an improved reflection type anti-

motion artifacts PPG detection model is proposed

based on electro-optic technique. Figure 1 shows the

structure of this novel design, the pulse wave

detection probe is constructed as two separate parts,

one part is a reflection type pulse detection module,

where two different wavelength LEDs and one PD

together composed as a sensor module, and this

module is electrically connected to the embedded

system. The other part is a motion artifact detection

module, where specifically a skin-friendly flexible

substrate with a narrow-band full-reflection film

plating on it is applied. Take a red light and green

light LED-PD sensor module for example, a flexible

substrate such as polydimethylsiloxane (PDMS) can

firmly adhere to the surface of the human body, and

the narrow-band full-reflection film is a red light full-

reflection film. The red light is totally reflected on the

surface of the film, and the green light can completely

transmit into the tissue theoretically. In this model,

the amount of red/green light received by PD can be

expressed as

normalR



(1)

bloodgnormal

QQQ

__gG





(2)

Equation (1) and (2) suppose that the red light

absorption

equals the amount of red light

absorbed by air(simplified), and the green light

absorption

is composed of the green light

absorbed by air(simplified) and human body AC

components. The thickness of the air layer for red and

green light is the same, that is, the optical path

difference introduced by the relative displacement

between the probe and the measured part is the same,

thus

normal

and

g_normal

Q are strongly correlated.

The red light signal is regarded as a reference signal

for noise, and the green light signal is regarded as a

mixed signal of the noise signal and the real signal, so

we can apply algorithms for motion artifact

cancellation.

2.2 The Proposed Embedded System

In this section, we present a wrist worn artifact-

resistive PWMP system, and the block diagram is

shown in Figure 2. The PPG signal is acquired by an

AFE (AFE4404 from Texas Instruments) after EMI

filter, then the signal is amplified and 24bit-ADC

sampled before sent to micro-controller through I

data communication protocol.

Model Design and System Implementation for the Study of Anti-motion Artifacts Detection in Pulse Wave Monitoring

103

Figure 1: The structure of the improved anti-motion

artifacts PPG detection model.

The MCU is an ultra low power, 2.4GHz RF device,

it contains a 32bit ARM cortex-M3 processor. The

standby current is 1uA, and the shutdown current of

the MCU is 100nA, which makes it suitable for the

design of wearable monitoring platform. An

integrated high efficiency dc-dc converter

BQ25015(from Texas Instruments) is designed to

operate directly from a single cell lithium battery, the

output voltage is adjustable from 0.7V to VBAT, this

convert is highly integrated charge and power

management device targeted at space-limited

Bluetooth applications. In this design, PWMP can

work almost 10 hours continuous even when powered

by a 40mAh rechargeable lithium battery, which

makes it suitable for long term health monitor. The

pulse wave data is then transferred to PC through

wireless module, and the data can be acquired in real

time at the sample rate of 100Hz. The main board of

the proposed PWMP system is in size of 20 *15mm

which is a miniaturized embedded system smaller in

sized than previous work (Zhou et al.,2016). This

kind of system has widely usage in physiological

parameter monitoring area, and it is well established

in other works (Zhang,2007).

Figure 2: Block diagram of the Embedded system.

3 SOFTWARE AND ALGORITHM

We propose a simulation method to verify the model

mentioned above. Light transport in optically dense

random media is characterized by multiple scattering,

but recent studies have demonstrated that the

diffusion approximation fails to fully describe back-

scattering/reflection and transmission characteristics.

While Monte-Carlo simulation is an effective

approach for modeling light transport in scattering

and absorbing media such as tissue (Guo et al.,2002).

3.1 Monte-Carlo Approach

Monte Carlo approach (Monte-Carlo, MC) is a

probabilistic statistical method widely used in the

physical field to deal with particle transport problems.

The basic idea is to first establish a probability model

or a stochastic process corresponding to the physical

process, and simulate the process through a series of

random numbers, and then calculate the statistical

characteristics of the parameters obtained by

observing or sampling the model or

process(Wang,Steven & Jacques,1992). In the

simulation of light transport, the process can be

described as: Firstly, describe the characteristics of

tissue by the following parameters: the thickness d,

the absorption coefficient ua the scattering coefficient

us, the anisotropy factor g, and the refractive index n.

ua is defined as the probability of photon absorption

per unit infinitesimal path length, and us is defined as

the probability of photon scattering per unit

infinitesimal path length. Then the light beam is

equivalent to a photon flow, in which each photon

randomly interacts with the tissue during random

walk, absorption and scattering, etc. The MC

simulation is to generate a random number to sample

these interactions (the walking step/the angle of

scattering). Finally, by sequentially emitting photons

and tracking the number of photons reflected,

transmitted, and absorbed, the results of photon

propagation in the tissue (reflectance, transmittance,

etc.) are statistically represented.

There are many approaches for MC simulation,

including analogue Monte-Carlo (AMC) and

variance reduction Monte-Carlo (VRMC). AMC is

the most direct description of real physical processes.

The weight of the photon is always maintained at 1

until it is completely absorbed or leaves the area of

interest and never returns. In order to get reliable

results, it is necessary to simulate a large number of

photons, which takes a long time to calculate.

Therefore, VRMC is generally used to improve the

AMC. In this method, the photon is given a certain

BIODEVICES 2020 - 13th International Conference on Biomedical Electronics and Devices

104

initial weight. After each interaction with tissue, the

weight of the photon is attenuated. When the weight

of the photon is less than the given thresholds, we can

believe that the transport of this photon has little

contribution to the statistical results, and then use

Russian roulette interrupt technology to give photons

a certain chance of survival. If the photon survives, it

continues to walk in tissue and if the photon dies, the

next photon will be emitted.

In this paper, the MC simulation program is

written based on the above process by MATLAB to

simulate the basic frame of the proposed PPG module

with the photon number set as 10

3.2 Dynamic Optical Model of Human

Skin Tissue

For a static optical model of skin tissue, the MC

simulation can successfully track the interaction

between photons and biological tissue at that moment.

Since photon emission time is at a scale of ps, the

photon emission process can be regarded as transient.

Therefore, as long as we build a dynamic optical

model of the skin tissue, that is, a model of the optical

properties of the skin tissue changes with time, and

use MC to simulate at an interval to model PPG signal

in time domain. In actual physical process, the

processor drives the LEDs to emit light at a certain

frequency, and the PD receives the optical signal,

which accords with the MC simulation process.

As the largest organ of human body, skin is a

complex multi-layered structure. In short, skin can be

divided into the epidermis layer, the dermis layer and

the subcutaneous tissue (Cheong et al.,1990; Faber et

al.,2004). The reflection type PPG detection is to

detect the volume change of the blood in the blood-

containing dermis layer. In this paper, the model of

human skin is divided into six layers: the epidermis

layer, the mastoid dermis layer, the upper layer of

blood vessels, the reticular dermis layer, the deep

blood vessels and the dermis layer. The upper layer

of blood vessels and the deep blood vessels are made

up of dermal tissue and blood in different proportions,

and blood is evenly distributed. Most studies have

shown that the upper blood vessels are composed of

90% dermal tissue and 10% blood, and the deep blood

vessels are composed of 10% dermal tissue and 90%

blood. The sixth layer of dermis is attached to the

subcutaneous fat. The pumping action of the heart

causes blood pressure to change periodically, and the

optical properties of skin vary with the blood content

of dermis. By establishing a relationship between

blood pressure and skin optical properties, a dynamic

optical model of human skin tissue can be established.

3.3 PPG Signal Modeling without

Motion Artifact

By combining the MC simulation with the skin tissue

dynamic model, PPG signal modeling in time domain

can be performed. First, Gaussian functions are used

to fit the waveform of a standard pressure pulse wave

with a period of 1s as the input of the model. The

pressure value is used for calculating the skin optical

model at that moment. The sampling is performed at

a frequency of 100Hz, that is, 100 MC simulations are

performed per second. Each simulation calculates the

optical parameters of the skin tissue, and emits a

photon packet of 10

photons with a wavelength of

540 nm green light. A detector is placed at a position

of 0.2 to 0.3 cm from the incident position, which

proves to be the best position for reflection detection,

and the total weight of the photons emitted at the

position is summed.

This process can be described as a simulation

without motion artifact, that is, the probe and the

measured part are closely attached without

considering the motion artifact introduced by the

movement of air layer.

3.4 PG Signal Modeling with Motion

Artifact

On the basis of the MC model above, it is easy to

model the anti-motion artifacts probe proposed in

Figure 1 by adding an air layer whose thickness

changes with time for simulating the relative

displacement between the probe and the measured

part. Green light detects the mixed signal of the real

pulse PPG signal and motion artifact, and interacts

with a seven-layer MC model (one air layer, six skin

layers). Red light only interacts with the air layer and

detects only motion artifact. In the MC simulation,

when the red light photon travels to the lower surface

of the air layer, it will be reflected without using a

random number to judge its path, so that the red light

full-reflection film can be simulated. By changing the

variation of the thickness of the air layer, different

kinds of motion can be simulated.

As shown in Figure 3, the noise signal can

represent the variation of the thickness of the air layer,

and the mixed signal is strongly disturbed by noise. It

is hard to identify the typical dicrotic notch of the

pulse wave signal in Figure 3(a), and it is almost

impossible to distinguish the pulse wave in Figure 3(b)

and Figure 3(c).

Model Design and System Implementation for the Study of Anti-motion Artifacts Detection in Pulse Wave Monitoring

105

(a) The thickness of the air layer changes as a half sine curve that

varies at a frequency of 1 Hz, simulating periodic low frequency

motions, such as a low frequency arm swing .

(b) The thickness of the air layer changes as a half sine curve that

varies at a frequency of 3 Hz, simulating periodic high frequency

motions, such as a high frequency arm swing.

10% duty cycle, simulating a sudden shake, such as typing.

Figure 3: The result of PPG signal modeling with Motion

artifact.

4 RESULTS AND DISCUSSIONS

4.1 Simulation Result

Adaptive filtering is a filtering method developed on

the basis of Wiener filtering. Least Mean Square

(LMS) adaptive algorithm is a method that

continuously adjusts the parameters of the adaptive

filter to minimize the mean square value of the error

between the output signal and the expected response

(Chan & Zhang,2002).

To meet the real world interference and sensor

inputs, in this paper, the mixed signals and noise

signals in Figure 3 are stretched and compressed in

time domain by interpolation, then randomly

combined as the input of the adaptive filter(mixed

signals and noise signals are sampled at the same

time). Figure 4 shows the simulation results.

In time domain, as the adaptive algorithm

converges, the correlation with motion artifact can be

removed and the PPG signal becomes distinguishable.

From the results above, the motion artifact

cancellation detection model proposed in this paper

provides good reference signal for adaptive algorithm

by extracting the motion artifact signal separate

simutaneously, which shows that the proposed model

has the potential to recover pulse wave signal for

pulse rate monitoring for the proposed motion

artifacts.

Figure 4: The result of adaptive algorithm.

4.2 PWMP System Test Result

The proposed platform PWMP was estimated by

Fluke Index2 Vital Signs Simulators (produced by

Fluke Corp.USA) at Zhejiang Institute of Medical

Device Testing. The measurement accuracy of pulse

rate by PWMP was within ±2 beats per minute(bpm)

at the range of 30bpm to 240bpm compared with the

output of Fluke Index2 in stationary situation.

To verify the model proposed, a 30*30*5mm

high reflection filter was customized, the filter

showed high transmission to visible light bands and

high reflection to other bands. Pulse signals acquired

from wrist (left hand) were recorded by designed

PWMP. Two young healthy subjects joined in this

experiment for one week and performed the tests two

times per day. For this type of study formal consent is

not required. Measurements were carried out under

certain temperature (20~27) and humidity

conditions (50%~60%) compensated by air

condition.

Three kinds of typical postures were performed,

measurements began when the subject was in a seated

BIODEVICES 2020 - 13th International Conference on Biomedical Electronics and Devices

106

(a) (b) (c)

Figure 5: Experimental testing process, (a)stationary,

(b)free arm swing, (c)typing.

position. Firstly, the subject was asked to sit in

stationary situation mode, see in Figure 5(a); Then,

free arm swing movement was applied, see in Figure

5(b); At last, typing for five characters was

applied ,see in Figure 5(c).

During the experiment, a short break for sixty

seconds was arranged between every two kinds of

motions. At the mean time, the PC-60B medical

Finger Clip Pulse Oximeter from Heal Force was

worn at the index finger of the subject's right hand to

work as criterion device. The right hand remained

stationary during the experiment, and the pulse rate

displayed on the pulse Oximeter was also recorded at

an interval of 5s as the criteria data.

Three kinds of typical signals of two channel PPG

lasted a period of 5s including green light PPG

channel(signal with motion accordingly), infrared

light PPG channel(motion accordingly) from wrist

were captured. Figure 6 demonstrated the typical

signal and signal process during free arm swing

movement. The blue dotted waveform (in the bottom

part of Figure 6, marked as noise) stood for noise was

highly consistent with the motion disturbance

waveform of infrared light(int the up part of Figure 6,

marked as ir). The red waveform, from which pulse

rate calculated and some maintain good

morphological characteristics of pulse wave, stood

for the result after signal process. In this case, the

calculated pulse rate was 91bpm compared with

89bpm form PC-60B medical Finger Clip Pulse

Oximeter.

Figure 7 demonstrated the typical signal and

signal process of typing. As infra(ir) channel showed,

the artifact during tying was random disturbance at

high frequencies. The signal obtained by green

channel was similar to the infra channel, the high

intensity of motion artifact overwhelmed the original

pulse wave shape. After signal process, the blue

dotted waveform stood for noise was highly

consistent with the motion disturbance waveform of

infrared light, and the red waveform, from which

pulse rate calculated, stood for the result after signal

process. In this case, the calculated pulse rate was

77bpm compared with 78bpm form PC-60B .

Figure 6: The typical signal and signal process of free arm

swing movement.

Figure 7: The typical signal and signal process of typing.

During the experiment, 1800 pieces of data were

collected, LMS along with Independent Component

Algorithm (ICA) was applied for signal processing.

Using PASW Statistics 18.0(SPSS Statistics)

software to analyze the data, we found that the

average and standard deviation of the pulse rate

obtained by proposed platform were 74.24 and 9.39

while they were 74.08 and 8.63 by using PC-60B. The

Pearson correlation coefficient was 0.953(p<0.01).

Model Design and System Implementation for the Study of Anti-motion Artifacts Detection in Pulse Wave Monitoring

107

The average deviation and standard deviation was

smaller than previous work as described (Zhou et al.,

2014). Obviously, the proposed platform has good

correlation as compared to the PC-60B in the

measurement of pulse rate.

The Bland-Altman plot of the PWMP calculated

in different postures and PC-60B in stationary

situation obtained were given in Figure 8. The x-axis

presented the average of the two methods, while the

y-axis showed the difference between them. We

observed that a total of 95.5% of the pulse rate

measurements lay in the limits of agreement (1.96 

SD), which indicated that the pulse rate calculated by

our proposed method was in close agreement with the

PC-60B in stationary accordingly.

Figure 8: Bland-Altman analysis of PWMP calculated in

different postures and PC-60B in stationary situation

accordingly.

In Table 1, error analysis in different postures

were presented. We could find that the mean square

error (MSE) was lowest in stationary posture, that

was mainly as in this posture, no movement was

applied. In free arm swing MSE had the largest value,

that was mainly because in this situation, pulse rate

changed during arm swing and more artifact was

applied into the platform. As typing last only a short

time, the influence was not so strong as free arm

swing.

By analyzing the waveform and comparing with

the pulse rate data collected by Pulse Oximeter, the

validity of the model was also verified from the

perspective of experiment. Zhang et al., (2019)

proposed a novel modular algorithm framework for

motion artifacts removal based on different

wavelengths for wrist-worn PPG sensors, and their

framework was most effective in removing artifacts

induced by micromotions from PPG signals using IR

PPG as a motion reference. Our proposed module was

tested under three kinds of typical postures(limited),

more motion scenes should be designed for

experimental verification in further works.

Table 1: Error analysis in different postures.

Typical

postures

PWMP

mean±SD

/bpm

PC-60B

mean±SD

/bpm

MSE

(Mean

square

error)

/bpm

Stationary 73.00±3.16 72.33±2.50 2.00

Free arm

swing

80.67±12.95 79.33±12.95 5.66

Typing 70.83±7.52 72.52±6.89 5.33

5 CONCLUSION

Unlike traditional pulse wave monitoring system

designs based on priori knowledge or through a lot of

experiments and measurements, this paper proposes a

commonly used simulation method and an embedded

system accordingly to study anti-motion artifacts

detection in pulse wave monitoring based on PPG. In

this model, a narrow-band full-reflection film plating

on a skin-friendly flexible substrate is specially

designed for extracting the motion artifact introduced

by air. Compared to the conventional MC simulation

for static skin model, a new method for PPG signal

modeling is also proposed, that is based on Monte-

Carlo approach to simulate the dynamic human skin

model, so as to achieve the PPG signal. Both PPG

signals with/without motion artifact are simulated and

the simulation results show that the motion artifact

cancellation detection model proposed in this paper

provides good reference signal for adaptive algorithm

by extracting the motion artifact signal separately.

The measurement accuracy of pulse rate by PWMP

meets the essential performance from the standard of

“YY 0784-2010”. To verify the model proposed,

three kinds of typical postures are performed and the

results show that the proposed platform has good

correlation as compared to the PC-60B in the

measurement of pulse rate and the proposed model

has the potential to recover pulse wave signal for

pulse rate monitoring.

The modeling method for PPG signal proposed in

this paper also can be used for further study. The

model and embedded system proposed here are

intended for continuous long-time, real-time pulse

wave monitoring. It has the potential application in

long-term human vital sign motoring, especially in

blood oxygen, pulse rate and PPG based blood

BIODEVICES 2020 - 13th International Conference on Biomedical Electronics and Devices

108

pressure monitoring. But there are still some

problems that should be improved in future work, for

example, more motion scenes should be designed for

experimental verification (Hibbing,Mantis,2018),

and in order to obtain more accurate morphological

parameters from pulse wave, the signal processing

algorithm should also be improved for different

motion scenes.

ACKNOWLEDGMENT

This work was supported by the National Key R&D

Program of China (No.2017YFF0210803), this

research was also funded by the Fundamental

Research Funds for the Central Universities

(No.2019FZA5015) and Applied Research Project of

Science and Public Welfare Technology in Zhejiang

Province (No.2017C33136).

REFERENCES

Zhou, C., Tu, C., Tian, J., Feng, J., Gao, Y., & Ye, X.

(2015). A low power miniaturized monitoring system

of six human physiological parameters based on

wearable body sensor network. Sensor Review, 35(2),

210-218.

Nogami, H., Inoue, R., Hayashida, Y., Ando, H., Ueno, T.,

& Sawada, R. (2018, January). Multifunctional optical

sensor module integrated optical micro displacement

sensor and its application to a photoplethysmographic

sensor with measuring contact force. In 11th

International Conference on Biomedical Electronics

and Devices, BIODEVICES 2018-Part of 11th

International Joint Conference on Biomedical

Engineering Systems and Technologies, BIOSTEC

2018, 71-76.

Davoudi, K., Shayeganyan, M., Tavakolian, K., &

Kaminska, B. (2014, March). Mechanically Flexiable

Biosensor for Detection of Photoplethysography. In

BIODEVICES, 159-164.

Zhou, C. C., Tu, C. L., Gao, Y., Wang, F. X., Gong, H. W.,

Lian, P., & Ye, X. S. (2014, September). A low-power,

wireless, wrist-worn device for long time heart rate

monitoring and fall detection. In 2014 International

Conference on Orange Technologies,IEEE,33-36.

Ra, H. K., Ahn, J., Yoon, H. J., Ko, J., & Son, S. H. (2016,

November). Accurately Measuring Heart Rate Using

Smart Watch. In Proceedings of the 14th ACM

Conference on Embedded Network Sensor Systems

CD-ROM, 380-381.

Tamura, T., Maeda, Y., Sekine, M., & Yoshida, M. (2014).

Wearable photoplethysmographic sensors — past and

present. Electronics, 3(2), 282-302.

Fukushima, H., Kawanaka, H., Bhuiyan, M. S., & Oguri, K.

(2012, August). Estimating heart rate using wrist-type

photoplethysmography and acceleration sensor while

running. In 2012 Annual International Conference of

the IEEE Engineering in Medicine and Biology

Society,IEEE, 2901-2904.

Zhou, C., Feng, J., Hu, J., & Ye, X. (2016). Study of

artifact-resistive technology based on a novel dual

photoplethysmography method for wearable pulse rate

monitors. Journal of medical systems, 40(3), 56.

Shimazaki, T., Hara, S., Okuhata, H., Nakamura, H., &

Kawabata, T. (2014, August). Cancellation of motion

artifact induced by exercise for ppg-based heart rate

sensing. In 2014 36th Annual International Conference

of the IEEE Engineering in Medicine and Biology

Society, IEEE, 3216-3219.

Zhang, G. (2007, August). A physical activities healthcare

system based onwireless sensing technology. In 13th

IEEE international conference on embedded and real-

time computing systems and applications (RTCSA

2007), IEEE, 369-376.

Guo, Z., Aber, J., Garetz, B. A., & Kumar, S. (2002). Monte

Carlo simulation and experiments of pulsed radiative

transfer. Journal of Quantitative Spectroscopy and

Radiative Transfer, 73(2-5), 159-168.

Wang, L., & Jacques, S. L. (1992). Monte Carlo modeling

of light transport in multi-layered tissues in standard C.

The University of Texas, MD Anderson Cancer Center,

Houston, 4-11.

Cheong, W. F., Prahl, S. A., & Welch, A. J. (1990). A

review of the optical properties of biological tissues.

IEEE journal of quantum electronics, 26(12), 2166-

2185.

Faber, D. J., Aalders, M. C., Mik, E. G., Hooper, B. A., van

Gemert, M. J., & van Leeuwen, T. G. (2004). Oxygen

saturation-dependent absorption and scattering of

blood. Physical review letters, 93(2), 028102.

Chan, K. W., & Zhang, Y. T. (2002, June). Adaptive

reduction of motion artifact from

photoplethysmographic recordings using a variable

step-size LMS filter. In SENSORS, 2002 IEEE,2,1343-

1346.

Bai, Y., Hibbing, P., Mantis, C., & Welk, G. J. (2018).

Comparative evaluation of heart rate-based monitors:

Apple Watch vs Fitbit Charge HR. Journal of sports

sciences, 36(15), 1734-1741.

Zhang, Y., Song, S., Vullings, R., Biswas, D., Simões-

Capela, N., Van Helleputte, N., Groenendaal, W.

(2019). Motion artifact reduction for wrist-worn

photoplethysmograph sensors based on different

wavelengths. Sensors, 19(3), 673.

Model Design and System Implementation for the Study of Anti-motion Artifacts Detection in Pulse Wave Monitoring

109