34th Annual International Conference of the IEEE EMBS

San Diego, California USA, 28 August - 1 September, 2012

A Clip-free Eyeglasses-based Wearable Monitoring Device for

Measuring Photoplethysmograhic Signals
Yali Zheng, Billy Leung, Stanley Sy, Yuanting Zhang and Carmen C.Y. Poon

Abstract— An eyeglasses-based device has been developed in signal processing techniques [6, 9] and optimizing the
this work to acquire photoplethysmogram (PPG) from the nose mechanical design [10]. However, motion artifact is still
bridge. This device is aimed to provide wearable physiological considered as one of the major challenges [11]. According to
monitoring without uncomfortable clips frequently used in PPG the knowledge of the author, the regular sites from which the
measurement from finger and ear. Switching control is applied PPG signal is measured include finger, earlobe and toe [12,
on the LED and photo detector for power saving. An experiment 13]. Few studies have considered the PPG signal from nose
involving postural change and treadmill jogging among 10 bridge.
healthy young subjects was carried out to evaluate the
performance of the device. Electrocardiogram (ECG) and PPG In this work, an eyeglasses-based PPG measurement
from finger, ear and nose were simultaneously recorded, from device is developed, which aims to provide stable, accurate
which heart rate (HR) and pulse transit time (PTT) were and comfortable monitoring of the physiological states of the
calculated. The results show that PPG measured from nose and subject. The sensing components are mounted on the frame of
ear are more resistant to motion than signal from finger during the eyeglasses and contact with the skin of the nose bridge. To
exercise. In addition, the difference between PTT measured evaluate the performance of the device, an experiment
from ear and nose indicates that local vasomotor activities may involving postural change and treadmill exercise was carried
exist on ear and/or nose channel, and suggests that PPG from out in this study.
different sites should be used for cuff-less PTT-based BP
estimation. We conclude that this wearable device has great
II. SYSTEM DEVELOPMENT
potential to be used in the healthcare management in the future.
I. INTRODUCTION A. System Architecture
Wearable monitoring technology has been considered as The eyeglasses-based PPG measurement device contains
one of the great challenges of health informatics to practice the following parts: optical sensor, sample-and-hold (S/H)
advancing healthcare management featured by low cost, high circuit, filter, micro-controller unit (MCU, ATmega 8),
efficiency and high quality [1]. Photoplethysmography is a bluetooth module (WT12) and cell phone or PC. The photo of
very popular and low-cost optical technique to detect the the eyeglasses-based device and the block diagram of the
device are shown in Fig. 1.
. This technique has been applied in
many clinical areas in physiological monitoring, such as
cuff-less and noninvasive BP estimation [3, 4], heart rate
monitoring [5], blood oxygen saturation (SpO2) monitoring
[6], and clinical assessment of some vascular diseases [7],
autonomic regulatory function [8].
Due to the portable nature of wearable monitoring devices,
two major issues which need to be considered in the design of
photoplethysmogram (PPG) measurement device are motion
artifacts and power consumption of the device. Motion artifact
is the major problem which constrains the practical
application of PPG measurement and many efforts have been
made to overcome this problem, such as adopting advanced

Manuscript received April 23, 2012. This work was supported by the Figure 1. The eyeglasses-based PPG measurement device and the block
Hong Kong Innovation and Technology Commission (ITS/159/11), the 973 diagram of the device
Project Fund (2010CB732606), and the Guangdong LCHT Innovation
Research Team Fund in China.
Yali Zheng, Stanley Sy and Yuanting Zhang are with the Department of
The PPG signal from nose is detected in reflective mode in
Electronic Engineering, The Chinese University of Hong Kong, Shatin, this device. The sensing components consist of a near-infrared
Hong Kong. LED emitter and a photo diode (PD), which are
Yuanting Zhang is also with the Key Laboratory for Health Informatics of
surface-mounted onto the nose bridge of the eyeglasses with a
Chinese Academy of Science (HICAS), Shenzhen, China.
Billy Leung and Carmen C.Y. Poon ([email protected]) are distance of 6 mm. The wavelength of the detected optical
with the Department of Surgery, The Chinese University of Hong Kong, signal by the PD ranges from 800 nm to 850 nm.
Hong Kong.
978-1-4577-1787-1/12/$26.00 ©2012 IEEE 5022
 Since the LED emitter is responsible for a great portion of analysis. The beat-to-beat HR and PTT were then both
total power consumption of the device, switching signal is resampled at 1 Hz for further processing.
applied on the LED and PD to reduce power. The switching The difference between PTT (ΔPTT) measured from the
frequency is designed to be 125 Hz and the pulse width of finger and other two sites, i.e., ΔPTTFinger-ear and ΔPTTFinger-nose,
switching signals for the LED and PD are 6ms and 2ms, were also calculated. The respiration-related high frequency
respectively, as shown in Fig. 2. components (>0.15Hz) of ΔPTT were removed to focus on
the low frequency variability, which is commonly understand
as the frequency band related to vasomotion.

IV. RESULTS
Excluding the Bluetooth module, the power consumption
of the eyeglasses-based device is about 40 mW.

Figure 2. The switching signals for LED and photo diode A. Finger, ear and nose PPG during jogging
It was found that for all subjects, the waveform of PPG
The pass band of the analog band-pass filters applied on the measured from the finger was corrupted severely during
PPG signal is from 0.5 to 15 Hz. ECG and PPG from ear and jogging, while PPG measured from the earlobe and nose
finger are acquired by another in-house-designed circuitry bridge were relatively stable. Segment of PPG from the three
which is not shown here. The same type of sensor was used for sites and ECG of one subject during jogging were shown in
measuring PPG at the three sites. Fig. 3. The red stars indicated the detected feature points.
All the acquired signals are then output to the MCU to ECG
-600
perform analog-to-digital conversion with 500 Hz sampling
ECG/mV
rate and 10-bit resolution. After processing by the MCU, the -800

signals are sent to cell phones or computers via the Bluetooth -1000
255 256 257 258 259 260 261 262 263 264 265
module with baud rate of 57600 Bd. All signals are received PPGf
and stored in the computer or cellphone for further processing. 800
PPGf/mV

700

III. EXPERIMENT 600

500
255 256 257 258 259 260 261 262 263 264 265

A. Experiment Procedure PPGe

800
The experiment was conducted in a quiet room at an
PPGe/mV

700
ambient temperature of 25oC. 10 healthy young subjects aged 600
27 ± 4 years old participated in the experiment. Upon arrival, 500
255 256 257 258 259 260 261 262 263 264 265
the subject was instructed to fill the form with enquiries of
PPGn
age, gender, weight, height, disease history and medication in
1000
recent days.
PPGn/mV

800
600
Before data recording, each subject was asked to rest for at 400
200
least 5 min. The subject was firstly asked to maintain sitting 255 256 257 258 259 260
Time/s
261 262 263 264 265

and standing postures each, for 2 minutes, and then was

instructed to jog on the treadmill (C956, Precor, USA) for 1 Figure 3. Typical data segment of one subject during jogging, where PPGf,
min at the speed of around 1.2 km/h. After the mild exercise, PPGe, PPGn are PPG measured from finger, ear and nose, respectively
the subject was asked to sit down to recover for 2 min. PPG
from right index finger, right earlobe and nose bridge were B. HR, PTT and ΔPTT during four states
recorded simultaneously together with ECG throughout the
Fig. 4 and Fig. 5 present the averaged HR and PTT among
experiment.
10 subjects throughout the experiment, respectively. It is
shown that HR and PTT can be calculated from PPGs
B. Data Analysis measured from earlobe and nose bridge throughout the
The acquired ECG and all PPGs were filtered by low-pass experiment. Compared with HR and PTT measured from
filter with cutoff frequency at 30 Hz and 16 Hz, respectively. earlobe and nose bridge, HR and PTT obtained from finger
The R-peak of ECG and the peak of the first derivative of PPG were only available when the subjects were stable.
(dPPG) were detected. The beat-to-beat heart rate (HR) was
calculated according to the detected feature points. Pulse Fig. 4 shows that HR from ear and nose demonstrate very
transit time (PTT) which has been considered as a potential similar changing trend with HR from ECG during four states.
surrogate of cuff-less and noninvasive measurement of blood It increased instantly at the onset of standing and jogging, and
pressure [3] was computed beat by beat as the time interval then returned back rapidly to baseline level at the beginning of
from R-peak of ECG to the peak of dPPG. Distorted PPG recovery.
waveform due to motion artifacts was manually removed and
the corresponding HR and PTT values were excluded from the
5023
 100
HR similar changing trend: significantly decreased during Stand1
90 Sit Stand Recovery
ECG
and jogging, while no significant change during Stand2 and
recovery (p<0.05).
bpm

80 Jog
70
0 50 100 150 200 250 300 350 400

TABLE I. MEANS AND STANDARD DEVIATIONS OF HEART RATE AND

100 Signal lost due to
motion artifacts Finger
PULSE TRANSIT TIME OF 10 SUBJECTS AT DIFFERENT STATES
90
bpm

80 ECG Finger Ear Nose

70
0 50 100 150 200 250 300 350 400 Sit 75 ± 4 75 ± 4 75 ± 4 75 ±4

100
HR Stand 82 ± 5 82 ± 5 82 ± 4 82 ± 4
Ear (bpm) Jog 90 ± 5 -- 90 ± 4 90 ± 4
90
Rec 75 ± 4 75 ± 5 75 ± 4 75 ± 5
bpm

80

70
Sit -- 216 ± 5 177 ± 4 186 ± 5
0 50 100 150 200 250 300 350 400
PTT Stand1 -- *207 ± 6 *170 ± 5 *177 ± 6
100 (ms) Stand2 -- 214 ± 5 178 ± 4 184 ± 5

90
Nose Jog -- -- *168 ± 5 *173 ± 8
Rec -- 217 ± 5 176 ± 5 182 ± 7
bpm

80
All data are presented as mean ± SD. * Student’s t-test compared to the
70
0 50 100 150 200 250 300 350 400 baseline level (p<0.05 is considered as significance).
Time/s

The averaged ΔPTTFinger-ear and ΔPTTFinger-nose among 10

Figure 4. The averaged HR calculated from ECG and PPG measured from subjects during the experiment are shown in Fig. 6. It can be
three sites among 10 subjects seen that the trend of ΔPTTFinger-ear was different from that of
ΔPTTFinger-nose.
As shown in Fig. 5, PTT measured from earlobe and nose
bridge showed consistent changing trend throughout the Differences between PTT measured from two sites
experiment, where it decreased suddenly at the beginning of 46
Finger-ear
standing, gradually went back to baseline level during the first 44 Finger-nose

30 sec during standing (Stand1), and maintained at a level 42

comparable to the baseline for the remaining one and a half 40

minute of standing (Stand2). PTT decreased acutely when
PTT difference/ms

38
start jogging and returned to baseline during recovery. 36

PTT 34
225 Signal lost due to
Finger
220 Stand1 motion artifacts 32
215
PTTf/ms

30
210
Stand2
205
28
200
Sit Jog Recovery
195 26
0 50 100 150 200 250 300 350 400 0 50 100 150 200 250 300 350 400 450
Time/s
185
Ear
180 Figure 6. The averaged ΔPTT between finger and other two sites among 10
subjects
PTTe/ms

175

170

165

160
0 50 100 150 200 250 300 350 400 V. DISCUSSION
195
Nose
Changes of PTT are often regarded as indicators of BP;
190

185
however, during postural change, the redistribution of blood
PTTn/ms

180 can also affect the measurement of PTT, which has to be

175
confirmed by further studies. The results of Fig. 4, Fig. 5 and
170

165
Table I showed that PPG measured from the nose bridge by
0 50 100 150 200 250 300 350 400
Time/s the proposed eyeglasses-based device showed better
performance than PPG from finger and the same performance
Figure 5. The averaged PTT calculated from ECG and PPG measured from as PPG from ear in terms of HR and PTT monitoring.
three sites among 10 subjects Therefore, this PPG measurement device has great potential to
be developed into wearable devices for physiological
Table I summarizes the statistics of the HR and PTT monitoring during daily activities in the future. Most
during each state among 10 subjects. The results of the study importantly, compared with typical PPG measurement
showed that the statistics of HR from PPG of different sites are devices developed in the past, the eyeglasses-based PPG
comparable to that measured from ECG. measurement device in our study is more comfortable and
On the other hand, the absolute values of PTT from three convenient for daily use, since it mounts the sensors on the
sites are different due to the different pulse transmission eyeglasses and does not need to introduce additional clips.
length. Compared with baseline, PTT from three sites showed
5024
 The signal quality of PPG measured from the nose bridge ACKNOWLEDGMENT
may be affected by the following factors: 1) vibration of the The authors are thankful to Y.P. Liang at JCBME for his
treadmill’s motor. We observed that the signal quality contribution on developing the software for signal displaying
decreased during jogging when the motor was on; and 2) and storage in the computer.
slipping of the eyeglasses. Unlike the clip used in PPG
measurement from ear where the sensor has very stable
contact with the skin, the eyeglasses on the nose may slip REFERENCES
down and cause artifacts, especially during exercise. This can [1] X. F. Teng, et al., "Wearable medical systems for p-health," IEEE
be corrected by adding a belt to fasten the eyeglasses. Reviews in Biomedical Engineering, vol. 1, pp. 62-74, 2008.
[2] J. Allen, "Photoplethysmography and its application in clinical
As aforementioned, in addition to the monitoring of HR physiological measurement," Physiological Measurement, vol. 28,
and PTT, another important potential application of PPG pp. R1-R39, 2007.
measurement is for continuous, noninvasive and cuff-less BP [3] C. C. Y. Poon and Y. T. Zhang, "Cuff-less and noninvasive
estimation, which has been reported in a number of studies [13, measurements of arterial blood pressure by pulse transit time," in 27th
14]. The accuracy of beat-to-beat BP estimation based on PTT Annual International Conference of the IEEE-EMBS, Shanghai,
2005, pp. 5877-5880.
is still the great challenge in this area, especially during [4] W. Chen, et al., "Continuous estimation of systolic blood pressure
dynamic states, and many efforts have been made to address using the pulse arrival time and intermittent calibration," Medical and
this challenge. Some studies pointed out that the effect of Biological Engineering and Computing, vol. 38, pp. 569-574, 2000.
cardiac pre-ejection period (PEP) should be removed from [5] V. K. Jayasree, et al., "A simple and novel integrated opto-electronic
PTT for BP estimation [15]. As an approximation, ear PTT system for blood volume pulse sensing and heart rate monitoring,"
was subtracted from finger PTT to represent the vascular International Journal of Optomechatronics, vol. 1, pp. 392-403,
2007.
transit time along the peripheral vessels [16]. However, the [6] Y. S. Yan and Y. T. Zhang, "An efficient motion-resistant method for
results as shown in Fig. 6 indicated that the differences wearable pulse oximeter," IEEE Transactions on Information
between ΔPTTFinger-ear and ΔPTTFinger-nose were probably Technology in Biomedicine, vol. 12, pp. 399-405, 2008.
affected by local vasomotor activities exist in the blood [7] E. Rosato, et al., "The different photoplethysmographic patterns can
vessels of the ear and/or nose. Since ΔPTT is different from help to distinguish patients with primary and sclerodermic raynaud
phenomenon," American Journal of the Medical Sciences, vol. 340,
site to site, ΔPTT calculated from either ear or nose branch
pp. 457-461, Dec 2010.
should be used with care for accurate central BP estimation [8] P. M. Middleton, et al., "Peripheral photoplethysmography variability
and multiple peripheral branches would be necessary for analysis of sepsis patients," Medical & Biological Engineering &
accurate estimation. Further studies are needed to explore the Computing, vol. 49, pp. 337-347, Mar 2011.
method by making use of the multichannel information to [9] J. A. C. Patterson and G. Z. Yang, "Ratiometric artifact reduction in
improve the accuracy of BP estimation. low power reflective photoplethysmography," IEEE Transactions on
Biomedical Circuits and Systems, vol. 5, pp. 330-338, Aug 2011.
[10] M. Z. Poh, et al., "Motion-tolerant magnetic earring sensor and
wireless earpiece for wearable photoplethysmography," IEEE
VI. CONCLUSION
Transactions on Information Technology in Biomedicine, vol. 14, pp.
This work presented a novel eyeglasses-based device 786-794, 2010.
which can measure PPG from the nose bridge. This device is [11] O. Such, "Motion tolerance in wearable sensors - the challenge of
motion artifact," in 29th Annual International Conference of the
suitable for daily use as no clips are needed. The results IEEE Engineering in Medicine and Biology Society, 2007, pp.
showed its ability of providing stable and accurate 1542-1545.
monitoring of HR and PTT during mild exercise. In addition, [12] G. S. H. Chan, et al., "Spontaneous fluctuations in the peripheral
photoplethysmographic waveform: Roles of arterial pressure and
the different results in difference between PTT measured
muscle sympathetic nerve activity," American Journal of
from two sites indicate that local vasomotor activities may Physiology-Heart and Circulatory Physiology, vol. 302, pp.
exist in the peripheral channels and suggest that PPG from H826-H836, Feb 2012.
multiple sites should be made use of in PTT-based BP [13] M. Wong, et al., "An evaluation of the cuffless blood pressure
estimation based on pulse transit time technique: A half year study on
estimation. Therefore, this device is promising to be used in
normotensive subjects," Cardiovascular Engineering, vol. 9, pp.
physiological monitoring for modern healthcare 32-38, 2009.
management. [14] Y. Chen, et al., "Continuous and noninvasive blood pressure
The subjects recruited in this study are young, healthy measurement: A novel modeling methodology of the relationship
subjects and therefore further studies are needed to confirm between blood pressure and pulse wave velocity," Annals of
Biomedical Engineering, vol. 37, pp. 2222-2233, 2009.
whether the results obtained here still stand among older
[15] J. Y. A. Foo, et al., "Evaluation of blood pressure changes using
subjects. In addition, a much lighter pair of eyeglasses will be vascular transit time," Physiological Measurement, vol. 27, pp.
designed in the future to reduce the motion artifacts resulted 685-694, 2006.
from the slipping of the eyeglasses on the nose bridge. [16] J. Proença, et al., "Is pulse transit time a good indicator of blood
Furthermore, the underlying relationship between central BP pressure changes during short physical exercise in a young
and the different vascular transit time measured from different population?," in Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, 2010, pp. 598-601.
peripheral branches will be further studied to explore
solutions to improve the accuracy of BP estimation.

5025