Research Article

International Journal of Distributed

Sensor Networks
2018, Vol. 14(8)
Intelligent health monitoring system Ó The Author(s) 2018
DOI: 10.1177/1550147718794318
based on smart clothing journals.sagepub.com/home/dsn

Chung-Chih Lin1,2, Chih-Yu Yang1, Zhuhuang Zhou3,4 and Shuicai Wu3

Abstract
In this study, we proposed an intelligent health monitoring system based on smart clothing. The system consisted of
smart clothing and sensing component, care institution control platform, and mobile device. The smart clothing is a
wearable device for electrocardiography signal collection and heart rate monitoring. The system integrated our pro-
posed fast empirical mode decomposition algorithm for electrocardiography denoising and hidden Markov model–based
algorithm for fall detection. Eight kinds of services were provided by the system, including surveillance of signs of life,
tracking of physiological functions, monitoring of the activity field, anti-lost, fall detection, emergency call for help, device
wearing detection, and device low battery warning. The performance of fast empirical mode decomposition and hidden
Markov model were evaluated by experiment I (fast empirical mode decomposition evaluation) and experiment II (fall
detection), respectively. The accuracy and sensitivity of R-peak detection using fast empirical mode decomposition were
96.46% and 98.75%, respectively. The accuracy, sensitivity, and specificity of fall detection using hidden Markov model
were 97.92%, 90.00%, and 99.50%, respectively. The system was evaluated in an elderly long-term care institution in
Taiwan. The results of the satisfaction survey showed that both the caregivers and the elders are willing to use the pro-
posed intelligent health monitoring system. The proposed system may be used for long-term health monitoring.

Keywords
Intelligent health monitoring system, smart clothing, empirical mode decomposition, hidden Markov model,
electrocardiography

Date received: 9 March 2018; accepted: 11 July 2018

Handling Editor: Kenneth Loh

Introduction
According to the World Health Organization,1 the
aging population will rise to nearly 25% of the global
population in 2050, and this number can even reach
33% for the developed countries. The issues of elderly
health care are becoming increasingly important,
because more and more health problems occur with
age, including heart disease, physical decline, and phys-
ical resilience decline. Therefore, developing intelligent
elderly health care techniques is of vital importance.
The fast growing of wearable devices has brought
their application in intelligent care, including smart
bands2,3 and smart clothing.4–6 The wireless transmis-
sion modes include radio frequency identification
1
Department of Computer Science and Information Engineering, College
of Engineering, Chang Gung University, Taoyuan, Taiwan
2
Division of Neurology, Linkou Medical Center, Chang Gung Memorial
Hospital and College of Medicine, Chang Gung University, Taoyuan,
Taiwan
3
College of Life Science and Bioengineering, Beijing University of
Technology, Beijing, China
4
Faculty of Information Technology, Beijing University of Technology,
Beijing, China
Corresponding author:
Chung-Chih Lin, Department of Computer Science and Information
Engineering, College of Engineering, Chang Gung University, 259, Wen-
Hwa 1st Road, Kwei-Shan, Taoyuan 33302, Taiwan.
Email: cclin@mail.cgu.edu.tw

Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
2 International Journal of Distributed Sensor Networks
Figure 1. Architecture of the proposed intelligent health monitoring system.
device (RFID),7,8 Bluetooth,9 and Bluetooth low energy
(BLE).10,11 Great progress has been made in the field of
intelligent care.12–21 Yao et al.17 developed a low-cost
interrogation system that was able to dynamically inter-
rogate the antenna sensor and wirelessly transmit the
acquired data to a smart device. Sundaravadivel et al.18
introduced a piezo-electric-based accelerometer sensor
design which helped in tracking the physical activities
of family and friends. With the three-dimensional (3D)
printing technique, Zhang and Amft19 proposed a per-
sonal fitted regular-look smart eyeglasses frames
equipped with bilateral electromyography recording to
monitor the activity of temporalis muscles, so as to
monitor chewing and eating. Daher et al.20 developed
smart tiles–based elder tracking and fall detection sys-
tem, which used pressure sensors and accelerometers
hidden under the smart tiles. Yao et al.21 presented a
compact antenna sensor interrogator for pressure sen-
sing; they developed a frequency-modulated continuous
wave generator to detect the resonant frequency of
antenna sensors, which encoded the pressure informa-
tion. However, an intelligent health monitoring system
based on smart clothing, which provides integrated ser-
vices such as cardiac function tracking and fall detec-
tion, is still lacking.
In this study, we designed and developed a smart
clothing–based intelligent health monitoring system,
which was composed of smart clothing and sensing com-
ponent, care institution control platform, and mobile
device. The smart clothing was used for electrocardiogra-
phy (ECG) signal collection and heart rate monitoring.
BLE was used for wireless data transmission.
The intelligent health monitoring system provided
eight kinds of services for the elders, including surveil-
lance of signs of life, tracking of physiological func-
tions, monitoring of the activity field, anti-lost, fall
detection, emergency call for help, device wearing
detection, and device low battery warning. The system
integrated our proposed Fast-EMD (empirical mode
decomposition) algorithm for ECG denoising and
HMM (hidden Markov model)-based algorithm for fall
detection. Experiments were designed to evaluate the
two proposed algorithms. Also, the proposed system
was applied and evaluated in a real scenario of elderly
health care.
This article is organized as follows. In section
‘‘Materials and methods,’’ methods are described.
Experiments and results are presented in section
‘‘Experiments and results.’’ In section ‘‘Discussion and
conclusion,’’ discussion and conclusions are given.
Materials and methods
System design
The architecture of the proposed intelligent health mon-
itoring system is illustrated in Figure 1. The system was
mainly composed of three parts: (1) smart clothing and
sensing component, responsible for receiving data; (2)
care institution control platform, responsible for health
data analysis and abnormality alert; and (3) mobile
device, responsible for displaying of data and warning
of abnormal events, and for providing mobile services.
The smart clothing collected ECG signals through four
electrode patches and then transferred the analog ECG
signals to the sensing component through conductive
fiber. The sensing component converted the analog
ECG signals into digital signals. The G-Sensor of the
sensing component was used for collecting the gravita-
tional acceleration (GC) data. The microprocessor unit
(MPU) of the sensing component analyzed digital ECG
signals and postures (GC) in real time using specific
algorithms. The ECG and posture data were then trans-
mitted through BLE, which were received by the BLE-
to-WiFi receiver in the environment. The BLE-to-WiFi
Lin et al.
Figure 2. Smart clothing and sensing component.
receiver transmitted the received data to the back-end
management platform for analysis, where the data were
finally converted into meaningful health data. The
health data were presented in Web mode to the mobile
device of the nurse station or the caregiver.
The architecture of smart clothing and sensing com-
ponent is shown in Figure 2. The design details of the
smart clothing are illustrated in Figure 3. The smart
clothing was made of conductive fiber. The four elec-
trode patches on the smart clothing were used for col-
lecting analog ECG signals, which were sent to the
sensing component. The analog-to-digital converter
(ADC) of the sensing component converted the analog
ECG signals into digital signals with a sampling fre-
quency of 250 Hz. The digital ECG signals were ana-
lyzed by the MPU of the sensing component to obtain
the health data, which were sent as broadcast packets
in a frequency of 1 Hz through BLE.
The proposed intelligent health monitoring system
provided the following eight kinds of services (Table 1):
1. Surveillance of signs of life. The heart rate of the
elderly is detected at any time, and alert is trig-
gered when the heart rate is above 140 Hz or
lower than 50 Hz.
2. Tracking of physiological functions. Using the
collected data of step number, monthly and
daily step number variation is analyzed. For a
specific physical state, it is determined whether
the amount of exercise corresponding to the
physical state is achieved.
3
3. Monitoring of the activity field. The position of
the elderly in the care institution is recorded reg-
ularly and analyzed to track whether the posi-
tion is varied or whether the elderly stays
indoors.
4. Anti-lost. It is detected by indoor positioning
whether the elderly gets close to the alert area
defined by the care institution. The system is
capable of detecting up to 10 m.
5. Fall detection. Falls are detected at any time. On
detecting a suspected fall incident, an emergency
call for help packet will be sent immediately.
6. Emergency call for help. The sensing component
has an emergency call for help button, which
can be used for real-time call for help when the
body is found to be abnormal.
7. Device wearing detection. If there is no sensing
component within the system architecture, the
required service cannot be reached, so it is
important to detect whether the elderly wears
the sensing component.
8. Device low battery warning. For the same reason
as (7), it is necessary to detect whether the sen-
sing component is in low battery.
ECG signal denoising and heart rate calculation
ECG signals are strongly affected by movement arti-
facts, which must therefore be effectively removed.22–24
To this purpose, we used empirical mode decomposition
4 International Journal of Distributed Sensor Networks

Figure 3. Design details of the smart clothing.

Table 1. Services provided by the proposed intelligent health monitoring system.

Service Content Warning threshold

Surveillance of signs of life
Tracking of physiological functions
Monitoring of the activity field
Anti-lost
Fall detection
Emergency call for help
Device wearing detection
Device low battery warning
Monitoring of the heart rate of bedridden
elders
Daily/monthly activity, total activity time,
times of getting out of bed
Recording of activity locations and ranges
To remind all the staff when specific elders
enter into the alert area
Reminder of getting out of bed at night,
warning of staying too long in the toilet
during the day or the night, warning of
falling down
To immediately learn about the location of
the emergency
Abnormal wearing of the device
Monitoring of device battery
\50 Hz or .140 Hz
1. Does the activity time during the day
reach the set standard?
2. Is the fall-sleep time and sleep time
normal?
Reminder for no activity in more than 1 h
Set according to different fields
1. More than 15 min in the toilet
2. Warning of getting out of bed at night
3. Immediate warning of falling down
Notified by the elders themselves
Whether the smart clothing and the
sensor are properly connected
Battery lower than 20%

(EMD)25 to filter noise. Previous studies have demon- X

N

strated that EMD can reduce motion artifact and base- x(t) = IMFj (t) + rN (t) ð1Þ
j=1
line wander of ECG signals.22–24 With EMD, the ECG
signal, x(t), was decomposed into N intrinsic mode func- The pseudo-code of the EMD algorithm is as
tions (IMFs) and residual, r(t) follows:
Lin et al. 5

Figure 5. Illustration of QRS complexes and RR interval of

ECG signals.

In this study, we improved the EMD algorithm to

speed up the algorithm for wearable devices. First, cal-
culation of SD in (3.6) was removed and the stopping
criterion SD(i)\e in (3.7) was replaced with i\iter,
where iter was the iteration threshold set according to
different states of motion. In this work, iter was experi-
mentally set at 14. Second, the stopping criterion in (6)
was replaced with the number of extrema in rj (t)  2
and j  order, where order was a preset iteration thresh-
old for j. In this work, order was experimentally set at
12. The improved EMD algorithm is called ‘‘Fast-
EMD.’’ The flow chart of Fast-EMD is shown in
Figure 4.
With Fast-EMD, the denoised ECG signal, y(t), was
obtained by summing up the 1st, 2nd, ..., Kth IMFs,
K  N

Figure 4. Flow chart of the proposed Fast-EMD algorithm. X

K
y(t) = IMFj (t) ð2Þ
j=1
1. j 1 (jth IMF)
2. rj21(t) x(t) // residual Using the denoised ECG signal y(t), a novel QRS
3. Extract the jth IMF complex waveform (Figure 5) morphology analysis
3.1. hj, i1 (t) rj1 (t), i 1; // i: number of algorithm called MWqrs proposed in our previous
iterations; work6 was employed to differentiate between QRS
3.2. Extract local maxima and minima of hj, i1 (t); complexes and artifacts. Three feature points were cal-
3.3. Compute upper envelope Uj, i1 (t) and lower culated after the algorithm detected a possible QRS
envelope Lj, i1 (t) by interpolating local max- complex. Heart rate was calculated by
ima and minima of hj, i1 (t), respectively;
60
3.4. Compute the average of Uj, i1 (t) and Lj, i1 (t): HR = ð3Þ
tint
mj, i1 (t) (Uj, i1 (t) + Lj, i1 (t))=2;
3.5. Update: hj, i (t) hj, i1 (t)  mj, i1 (t), i i + 1; where the RR interval, tint, was the time between two
3.6. Calculate stopping criterion: adjacent R-peaks of the denoised ECG signal y(t), as
Pt
jhj, i1 (t)hj, i (t)j
2
shown in Figure 5.
SD(i) = 2 ;
t=0 ðhj, i1 (t)Þ
3.7. Decision: repeat (3.2) to (3.6) until SD(i)\e
and then put IMFj (t) hj, i (t); // jth IMF Fall detection
4. Update residual: rj (t) rj1 (t)  IMFj (t); In this study, a fall detection method based on accelera-
5. Repeat (3) with j j + 1; tion data and HMM proposed by our group was
6. Stop when the number of extrema in rj (t)  2. employed.26 For completeness, the method is described
6 International Journal of Distributed Sensor Networks

Figure 6. Four states of a fall.

in brief as follows. Accelerations along x, y, and z axes

are denoted as ax, ay, and az, respectively. The resultant
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
acceleration norm is A = a2x + a2y + a2z . The fall state
will be reflected in the change of body acceleration. We
can classify four distinct patterns of the resultant accel-
eration A as shown in Figure 6:

1. Balanced state. The amplitude of the resultant

acceleration curve is approximately 1 gravita-
tional (g) acceleration.
2. Imbalanced state. The acceleration is approxi-
mately 0g when the body loses balance in a short
period of time.
3. Falling state. There are multiple peaks once the
body collides with ground. The peak of the
resultant acceleration can exceed 4g.
4. Normal state. A person cannot stand up imme-
diately after a fall occurs and the acceleration is
almost 1g again.
The different states of resultant acceleration are col-
lected to build an HMM. We used the Baum–Welch (B-
W) algorithm27,28 to train the HMM, which involved
the following three steps:
1. Intercept the acceleration sampling data of
about 0.5 s with a time window to obtain a data
set {ai}, i = 1, 2, ..., 100. The data set {ai} is
divided into 10 data units, and the average value
of the sampled data in each data unit was calcu-
lated to obtain the time sequence {oi}, i = 1, 2,
..., 10.
2. In order to distinguish the different grades of
the motion state, the variation range of the
acceleration sampling data is divided into sev-
eral sections, and the feature values of each sec-
tion are defined to obtain the observing feature
sequence {Oi}, i = 1, 2, ..., 10. The
Figure 7. Flow chart of the proposed hidden Markov model
(HMM)-based fall detection method.
transformation from {oi} to {Oi} is as follows
8
> 1 if 0  oi  0:4 g
>
>
>
< if 0:4\oi  0:7 g
2
Oi = 3 if 0:7\oi  1 g , i = 1, 2, :::, 10 ð4Þ
>
>
>
> 4 if 1\oi \2 g
:
5 if 2 g\oi
3. The classical B-W algorithm is used for training
the HMM. The major idea of the B-W algo-
rithm is to update the probability weight of the
state by recursive iteration, so that the model
parameters can better explain the training
sequence. We set the threshold of probability P
at 51.4%. The probability threshold acts as the
criterion to detect the probability of a fall. The
flow chart of detecting falls is shown in Figure 7.
Experiments and results
Two groups of experiments were performed, experi-
ment I: Fast-EMD evaluation and experiment II: fall
detection. Ten young college students (male) were
Lin et al. 7

recruited (aged between 22 and 24, height ranging from Table 2. Accuracy of R-peak detection with the proposed fast
165 to 178 cm, and weight between 51 and 76 kg) to empirical mode decomposition (Fast-EMD) method.
participate in the two groups of experiments. Finally,
K 0 km/h 1 km/h 2 km/h 3 km/h
the proposed intelligent health monitoring system was
applied and evaluated in a real scenario. 1 22.38% 28.96% 23.81% 28.13%
2 94.87% 95.67% 92.49% 89.71%
3 97.42% 97.27% 96.22% 93.59%
Experiment I: Fast-EMD evaluation 4 97.70% 97.26% 96.53% 93.39%
5 97.76% 97.81% 96.33% 93.08%
The ECG data of the participants were collected as
6 97.71% 97.86% 96.67% 93.33%
they exercised for 5 min. They exercised at different 7 97.81% 97.61% 96.40% 92.98%
speeds such as 0, 1, 2, and 3 km/h. Each condition was 8 97.85% 97.68% 96.49% 92.54%
repeated twice. We compared the difference using the 9 97.76% 97.78% 96.44% 92.61%
data with and without Fast-EMD. The accuracy and 10 97.73% 97.68% 96.77% 92.82%
11 97.95% 97.80% 96.77% 92.52%
sensitivity of R-peak detection with Fast-EMD are 12 97.66% 97.53% 96.34% 92.24%
shown in Tables 2 and 3. From Tables 2 and 3, it can 13 97.78% 97.33% 96.21% 92.48%
be found that for speed  2 km/h, using K = 11 pro- 14 97.71% 95.84% 95.41% 91.12%
duced the best performance for R-peak detection; and
for speed = 3 km/h, using K = 10 did. If we cannot
predict the speed, using K = 11 will generally produce
the best performance, with the overall accuracy and Table 3. Sensitivity of R-peak detection with the proposed fast
sensitivity of R-peak detection being 96.46% and empirical mode decomposition (Fast-EMD) method.
98.75%, respectively.
K 0 km/h 1 km/h 2 km/h 3 km/h

Experiment II: fall detection 1 77.85% 82.84% 78.94% 81.32%

2 99.41% 99.63% 98.68% 98.19%
We collected 150 datasets pertaining to normal activi- 3 99.21% 99.45% 98.71% 97.90%
ties including walking, jogging, sitting, standing, and 4 99.41% 99.45% 98.76% 97.54%
5 99.42% 99.67% 98.56% 97.26%
climbing downstairs, and 150 sets of falling data includ- 6 99.39% 99.67% 98.83% 97.34%
ing falling forward, falling backward, falling to the side, 7 99.44% 99.56% 98.66% 97.09%
falling when walking, and falling when jogging. Cross 8 99.47% 99.58% 98.76% 96.65%
validation was used. The HMM was built using 30 sets 9 99.41% 99.62% 98.71% 96.87%
of fall data selected randomly. We built four HMMs 10 99.39% 99.58% 98.87% 97.00%
11 99.55% 99.56% 98.91% 96.87%
using 120 sets. The remaining 30 falling data and 150 12 99.34% 99.47% 98.64% 96.63%
normal activities’ data were then used as testing data. 13 99.44% 99.47% 98.68% 96.69%
The performance of the HMM-based fall detection 14 99.39% 98.70% 98.22% 95.98%
method is shown in Table 4. The average accuracy, sen-
sitivity, and specificity of fall detection using the HMM
were 97.92%, 90.00%, and 99.50%, respectively.
Table 4. Performance of the proposed hidden Markov model
(HMM)-based fall detection method.
Application in real scenario
# Accuracy Sensitivity Specificity
The proposed intelligent health monitoring system has
been applied in an elderly long-term care institution, 1 97.78% 93.33% 98.67%
that is, Yi-De Elderly Long-term Care Center, Min- 2 98.89% 93.33% 100.00%
Sheng Healthcare, Taoyuan, Taiwan. The elderly care 3 98.33% 90.00% 100.00%
4 96.67% 83.33% 99.33%
institution has a room area of 9917 m2. The capacity is Average 97.92% 90.00% 99.50%
200 people. A floor of the institution was used for
experiment scenario, which was shaped like P. There
were 50 subjects from 10 rooms, with 5 subjects from information of the subjects, and altering when an emer-
each room. Each subject wore a set of smart clothing gency event occurred; the caregiver could check at any
with sensing components. time on the Web through the mobile device. In the
The nurse station at the floor was equipped with a environment, we used power over Ethernet (POE)-pow-
central control center. The main functions of the con- ered BLE-to-WiFi receivers. Each room was equipped
trol center included (1) adding, modifying, and deleting with one BLE-to-WiFi receiver. There were also a
personal basic data, and setting the BLE-to-WiFi recei- number of BLE-to-WiFi receivers equipped at walk-
ver; and (2) monitoring the position and gait ways, halls, elevators, and escape exits.
8 International Journal of Distributed Sensor Networks
Through the sensing component, the information of
subjects’ positions and gaits and component abnormity
can be obtained. When a subject has an emergency,
such as falling, entering into the alert area (elevator,
escape exit, etc.) and pressing the emergency button on
the sensing component, the caregiver can receive the
alert notification and conduct real-time treatment
through the central control center at the nurse station
or through the mobile device.
The proposed intelligent health monitoring system
collected and recorded the subjects’ data of daily events,
including the times of staying in situ for over 2 h, times
of approaching the alert area, times of falls, times of
getting out of bed at night, times of staying in the toilet
for over 15 min, times of emergency call for help, and
sleep time and activities. The system conducted data
analysis based on these indices to assist the caregiver in
easily understanding the status of each subject, so as to
improve the quality of care.
Satisfaction survey of the proposed intelligent health
monitoring system was conducted to the caregivers and
the subjects. The survey results showed that 95% of the
caregivers would like to use the system, 97% of the care-
givers thought that the system was easy to operate, 92%
of the subjects would like to use the system, and 98% of
the subjects thought that the smart clothing was com-
fortable and was willing to wear the smart clothing for a
long time. In the future, the proposed intelligent health
monitoring system will be applied and improved for
healthy communities and home-based elderly care to
achieve the purpose of elderly care in situ.
Discussion and conclusion
We have proposed a smart clothing–based intelligent
health monitoring system. The system is mainly com-
posed of smart clothing and sensing component, care
institution control platform, and mobile device. The
system can provide eight kinds of services for the elders,
including surveillance of signs of life, tracking of phy-
siological functions, monitoring of the activity field,
anti-lost, fall detection, emergency call for help, device
wearing detection, and device low battery warning. To
our knowledge, such a system has not been reported in
the literature.
The system has integrated our proposed Fast-EMD
algorithm for ECG denoising and the HMM algorithm
for fall detection. The Fast-EMD algorithm was
improved based on the EMD algorithm for wearable
devices to suppress motion artifacts. The HMM-based
fall detection algorithm was originally proposed in our
previous study.26 However, the algorithm was imple-
mented and evaluated in an off-line way in the study by
Cao et al.26 Compared with the study by Cao et al.,26
the improvements in this work were as follows: (1) we
implemented the HMM-based fall detection algorithm
in the wearable device and (2) the algorithm was evalu-
ated in an online manner through the proposed intelli-
gent health monitoring system. The performance of
the Fast-EMD and HMM algorithms have been evalu-
ated by experiment I and experiment II, respectively.
The evaluation results demonstrate the satisfying
performance of the proposed Fast-EMD and HMM
algorithms.
In recent work,29 a remote health monitoring system
for the elderly based on smart home gateway was intro-
duced, which consisted of three parts: smart clothing,
smart home gateway, and health care server. Compared
with the work by Guan et al.,29 we have achieved the
following improvements in this work. The smart
clothing–based intelligent health monitoring system in
this work integrated our proposed Fast-EMD algo-
rithm for ECG signal denoising and the HMM-based
fall detection algorithm, supporting some of the eight
kinds of services provided by the system. Moreover, the
proposed system was applied and evaluated in a real
elderly care institution, with satisfying performance.
In conclusion, we have designed and developed an
intelligent health monitoring system based on smart
clothing in this work. A Fast-EMD-based ECG denois-
ing method and an HMM-based fall detection method
have been proposed and integrated into the system.
The system can provide eight kinds of services to the
elderly. The proposed intelligent health monitoring sys-
tem has been applied and evaluated in a real scenario,
that is, an elderly long-term care institution. The results
of the satisfaction survey show that both the caregivers
and the elders are willing to use the system. The future
direction of the proposed intelligent health monitoring
system is elderly care in situ for healthy communities
and home-based elderly care.
Acknowledgements
The authors would like to thank the anonymous reviewers for
their insightful comments and suggestions.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) disclosed receipt of the following financial sup-
port for the research, authorship, and/or publication of this
article: This work was supported by Ministry of Science and
Technology (National Science Council) of Taiwan (Grant
Nos. MOST 104-2218-E-182-005-MY3 and MOST 105-2627-
E-182-001) and Chang Gung Memorial Hospital (Grant No.
CMRPD3E0063). S.W. was supported by National Natural
Lin et al.
Science Foundation of China (Grant No. 71661167001). Z.Z.
was supported by Beijing Natural Science Foundation (Grant
No. 4184081), China Postdoctoral Science Foundation
(Grant No. 2017M620566), Postdoctoral Research Fund of
Chaoyang District, Beijing (Grant No. 2017ZZ-01-03), and
Basic Research Fund of Beijing University of Technology.
References
1. World Health Organization. World report on ageing and
health, 2015, http://www.who.int/ageing/events/world-
report-2015-launch/en/ (accessed 30 March 2018).
2. Nguyen ND, Truong PH and Jeong GM. Daily wrist
activity classification using a smart band. Physiol Meas
2017; 38(9): L10–L16.
3. Eun SJ, Whangbo TK, Park DK, et al. Development of
personalized urination recognition technology using
smart bands. Int Neurourol J 2017; 21(Suppl 1): S76–S83.
4. Axisa F, Schmitt PM, Gehin C, et al. Flexible technolo-
gies and smart clothing for citizen medicine, home
healthcare, and disease prevention. IEEE Trans Inf Tech-
nol Biomed 2005; 9(3): 325–336.
5. Chen M, Ma Y, Song J, et al. Smart clothing: connecting
human with clouds and big data for sustainable health
monitoring. Mobile Netw Appl 2016; 21(5): 825–845.
6. Wang J, Lin CC, Yu YS, et al. Wireless sensor-based
smart-clothing platform for ECG monitoring. Comput
Math Methods Med 2015; 2015: 295704.
7. Zhang J, Tian GY, Marindra AMJ, et al. A review of
passive RFID tag antenna-based sensors and systems for
structural health monitoring applications. Sensors 2017;
17(2): 265.
8. Aslam B, Khan UH, Azam MA, et al. Miniaturized
decoupled slotted patch RFID tag antennas for wearable
health care. Int J RF Microw C E 2017; 27(1): e21048.
9. Vazquez M, Garibaldi JA, Nieto JI, et al. Model for per-
sonalization of mobile health systems for monitoring
patients with chronic disease. IEEE Lat Am Trans 2016;
14(2): 965–970.
10. Mahmud MS, Wang H, Esfar E, et al. A wireless health
monitoring system using mobile phone accessories. IEEE
Internet Things J 2017; 4(6): 2009–2018.
11. Hussein AF, Hashim SJ, Aziz AFA, et al. A real time
ECG data compression scheme for enhanced Bluetooth
low energy ECG system power consumption. J Ambient
Intell Human Comput. Epub ahead of print 17 August
2017. DOI: 10.1007/s12652-017-0560-y.
12. Pigini L, Bovi G, Panzarino C, et al. Pilot test of a new
personal health system integrating environmental and
wearable sensors for telemonitoring and care of elderly
people at home (SMARTA project). Gerontology 2017;
63(3): 281–286.
13. Majumder S, Aghayi E, Noferesti M, et al. Smart homes
for elderly healthcare—recent advances and research
challenges. Sensors 2017; 17(11): 2496.
14. Wang Z, Yang Z and Dong T. A review of wearable tech-
nologies for elderly care that can accurately track indoor
position, recognize physical activities and monitor vital
signs in real time. Sensors 2017; 17(2): 341.
15. Kekade S, Hseieh CH, Islam MM, et al. The usefulness
and actual use of wearable devices among the elderly
9
population. Comput Methods Programs Biomed 2018;
153: 137–159.
16. Baig MM, GholamHosseini H, Moqeem AA, et al. A
systematic review of wearable patient monitoring
systems—current challenges and opportunities for clinical
adoption. J Med Syst 2017; 41(7): 115.
17. Yao J, Skilskyj J and Huang H. A compact, low-cost,
real-time interrogation system for dynamic interrogation
of microstrip patch antenna sensor. In: Proceedings of the
SPIE smart structures and materials nondestructive evalua-
tion and health monitoring, Denver, CO, 27 March 2018,
vol. 10598, p.105980Y. Bellingham, WA: SPIE.
18. Sundaravadivel P, Mohanty SP, Kougianos E, et al.
Smart-walk: an intelligent physiological monitoring sys-
tem for smart families. In: Proceedings of the IEEE inter-
national conference on consumer electronics (ICCE), Las
Vegas, NV, 12–14 January 2018, pp.1–4. New York:
IEEE.
19. Zhang R and Amft O. Monitoring chewing and eating in
free-living using smart eyeglasses. IEEE J Biomed Health
Inform 2018; 22(1): 23–32.
20. Daher M, Diab A, El Najjar MEB, et al. Elder tracking
and fall detection system using smart tiles. IEEE Sensors
J 2016; 17(2): 469–479.
21. Yao J, Tjuatja S and Huang H. A compact FMCW inter-
rogator of microstrip antenna for foot pressure sensing.
In: Proceedings of the progress in electromagnetic research
symposium (PIERS), Shanghai, China, 8–11 August
2016, pp.2101–2105. New York: IEEE.
22. Blanco-Velasco M, Weng B and Barner KE. ECG signal
denoising and baseline wander correction based on the
empirical mode decomposition. Comput Biol Med 2008;
38(1): 1–13.
23. Jain S, Bajaj V and Kumar A. Riemann Liouvelle frac-
tional integral based empirical mode decomposition for
ECG denoising. IEEE J Biomed Health Inform 2018; 22:
1133–1139.
24. Lee J, McManus DD, Merchant S, et al. Automatic
motion and noise artifact detection in Holter ECG data
using empirical mode decomposition and statistical
approaches. IEEE Trans Biomed Eng 2012; 59(6):
1499–1506.
25. Huang NE, Shen Z, Long SR, et al. The empirical mode
decomposition and the Hilbert spectrum for nonlinear
and non-stationary time series analysis. P Roy Soc A:
Math Phy 1998; 454(1971): 903–995.
26. Cao H, Wu S, Zhou Z, et al. A fall detection method
based on acceleration data and hidden Markov mode. In:
Proceedings of the IEEE international conference on signal
and image processing (ICSIP), Beijing, China, 13–15
August 2016, pp.684–689. New York: IEEE.
27. Rabiner LR. A tutorial on hidden Markov models and
selected applications in speech recognition. Proc IEEE
1989; 77(2): 257–286.
28. Duda RO, Hart PE and Stork DG. Pattern classification.
Hoboken, NJ: John Wiley & Sons, 2012.
29. Guan K, Shao M and Wu S. A remote health monitoring
system for the elderly based on smart home gateway. J
Healthc Eng 2017; 2017: 5843504.