Sensors 2015, 15, 11295-11311; doi:10.

3390/s150511295
OPEN ACCESS

sensors
ISSN 1424-8220
www.mdpi.com/journal/sensors
Article

Heart Rate Variability Monitoring during Sleep Based on

Capacitively Coupled Textile Electrodes on a Bed
Hong Ji Lee 1, Su Hwan Hwang 1, Hee Nam Yoon 1, Won Kyu Lee 1 and Kwang Suk Park 2,*

1
Interdisciplinary Program for Bioengineering, Graduate School, Seoul National University,
Seoul 110-799, Korea; E-Mails: hongjidan@bmsil.snu.ac.kr (H.J.L.); lostzoo@bmsil.snu.ac.kr (S.H.H.);
hnyoon@bmsil.snu.ac.kr (H.N.Y.); wongyu86@bmsil.snu.ac.kr (W.K.L.)
2
Department of Biomedical Engineering, College of Medicine, Seoul National University,
Seoul 110-799, Korea

* Author to whom correspondence should be addressed; E-Mail: pks@bmsil.snu.ac.kr;

Tel.: +82-2-2072-3135; Fax: +82-2-3676-2821.

Academic Editor: Panicos Kyriacou

Received: 17 March 2015 / Accepted: 8 May 2015 / Published: 14 May 2015

Abstract: In this study, we developed and tested a capacitively coupled electrocardiogram

(ECG) measurement system using conductive textiles on a bed, for long-term healthcare
monitoring. The system, which was designed to measure ECG in a bed with no constraints
of sleep position and posture, included a foam layer to increase the contact region with the
curvature of the body and a cover to ensure durability and easy installation. Nine healthy
subjects participated in the experiment during polysomnography (PSG), and the heart rate
(HR) coverage and heart rate variability (HRV) parameters were analyzed to evaluate the
system. The experimental results showed that the mean of R-peak coverage was 98.0%
(95.5%–99.7%), and the normalized errors of HRV time and spectral measures between the
Ag/AgCl system and our system ranged from 0.15% to 4.20%. The root mean square errors
for inter-beat (RR) intervals and HR were 1.36 ms and 0.09 bpm, respectively. We also
showed the potential of our developed system for rapid eye movement (REM) sleep and
wake detection as well as for recording of abnormal states.

Keywords: capacitively coupled electrodes; conductive textiles; ECG; HRV monitoring

Sensors 2015, 15 11296

1. Introduction

With the population aging around the world, interest in leading a healthy life rather than undergoing
post-sickness treatment has been rising. To realize this, long-term home healthcare systems that can
detect abnormal signs in healthy people at an early stage and observe the progress of disease in
non-hospitalized patients have been developed. However, taking manual measurements of
physiological signals several times a day is inconvenient to people. Therefore, it is important to obtain
valuable information on health condition every day without interfering with daily activity.
Unconstrained and unconscious home-based daily health monitoring systems measuring
electro-cardiograms (ECGs), ballistocardiograms (BCGs), phonocardiograms (PCGs), and
photo-plethysmograms (PPGs) have been developed by various researchers [1–20]. However, such
non-contact sensing systems are vulnerable to motion artifacts. Therefore, considering that we spend a
third of our lives sleeping and make relatively few movements during this period, the most suitable
approach is long-term health monitoring during sleep. This approach would also minimize the
intrusion on daily activities and enable accumulation of a large amount of data.
In health monitoring systems, ECG signals are most widely used for observing heart or
cardiovascular functions by beat-to-beat analysis because mechanical signals are not as sharp and
clear [16]. The conventional ECG measurement systems require wired and wet adhesive Ag/AgCl
electrodes, but this is not suitable for daily long-term monitoring because of the use of disposal
electrodes, skin irritation from direct contact electrodes, and limitation of activity caused by the wired
system. Therefore, several researchers have proposed dry electrodes and non-contact electrodes [21].
However, dry electrodes could lead to metal allergy due to direct contact with bare skin [19]. To
overcome these limitations, capacitively coupled electrodes that use non-contact sensing technology
have been developed [22–24].
Several studies have presented unconstrained ECG measurement systems using dry or capacitive
sensing methods in a bed during sleep. Ishijima [13] and Devot et al. [14] proposed an ECG
monitoring system with textile electrodes located on a foot mat and pillow. However, because this
system requires direct contact with the bare skin of the neck and foot, there was a contact problem
when sleeping with legs bent or sleeping outside the pillow. In addition, since the textile was attached
on the entire pillow, the ECG signals were greatly influenced by noise caused by hair. Park et al. [15]
presented a system with two electrode sets made of conductive textiles located on the pillow, leg, and
shoulders. The ECG signals were measured from the two electrodes in direct contact with skin using a
channel selection algorithm. Another approach to measuring ECG with direct skin contact sensors was
introduced by Peltokangas et al. [16]. Eight embroidered fabric electrodes were sewn on a bed sheet,
and ECG was recorded from one contacted bipolar channel among seven channels. These ECG
measurement systems using direct skin contact dry electrodes [13–16] are complicated to set-up and
removing because the textile electrodes are attached directly to a mattress or the bedcover and
pillowcase. In addition, there is a high risk of contamination due to external exposure of electrodes.
Lim et al. [17] proposed a non-contact ECG measurement system using an array of eight copper
electrodes inserted in a mattress and a large textile electrode. However, because the rigid and hard
electrodes protruded from the mattress, the user’s back could become sore when they slept on a bed for
a long time. Moreover, the systems developed by Peltokangas et al. [16] and Lim et al. [17] need a
Sensors 2015, 15 11297

channel selection method because they use multi-channel electrodes, and they might miss
better-quality signals in the process of selecting contacted electrodes.
Three groups of researchers designed electrode configurations with wide strips based on conductive
textiles for ECG measurement in a bed without direct skin contact [18–20]. Wu and Zhang [18]
additionally used a leading tail beside each electrode sewn beneath the bed sheet to connect a pre-amp
module. However, because the width of electrode (50 cm) covered only the middle part of the mattress,
ECG signals could not be recorded when lying on the edge of the bed or on the edge of the electrodes
in a lateral posture. Ueno and Yama [19] and Ishida et al. [20] proposed a bed-sheet unit, but it had
some limitations in set-up and removing the sensors because the rectangular conductive fabric was
stuck to the bed sheet with adhesive. In addition, the systems also had a high risk of contamination and
a high chance of contacting the electrodes directly with hands, as the textiles were exposed to the
external environment.
Regardless of the sensing technology used, the systems mentioned above can provide meaningful
information about health, such as acute mental stress, sleep quality, and cardiac arrhythmias (sinus
arrhythmia, bradyarrhythmia, Premature Atrial Contractions, Multifocal atrial tachycardia, supraventricular
tachycardia, and so on), by analyzing R-peak intervals and heart rate variability (HRV) during
sleep [25–28]. However, there have been very few studies on evaluating the HRV measures in time
and frequency domains measured from a capacitively coupled ECG sensing system.
In clinical research, sleep fragmentation due to nocturnal awakenings can lead to daytime
sleepiness, obesity, diabetes, cognitive dysfunction, worsening cardiovascular risk, and impairment of
immune functions [29,30]. In addition, rapid eye movement (REM) sleep behavior disorder and
narcolepsy happen in REM stage [31]. Therefore, it is important to monitor the frequency of nocturnal
awakenings and the durations of wakefulness and REM during sleep. The golden standard for sleep
monitoring is polysomnography (PSG), but PSG is no guarantee of observing the usual sleep condition
because people sleep in an unfamiliar environment with multiple wired sensors attached on the face,
head, and limbs. Moreover, there are few clinical centers that install PSG system, and the cost is quite
expensive. Therefore, it is a better approach to use unconstrained systems for continuous sleep and
cardiovascular monitoring at home.
In this study, we developed a capacitively coupled ECG recording system with conductive textiles
for unconstrained daily health monitoring on a bed at home. The main contributions of the proposed
system are: (1) ECG was measured without any direct contact between the body and electrodes; (2) the
array and size of sensing electrodes were designed to measure ECG with no constraints of sleep
posture and location in a bed; (3) polyurethane foam layer was used to minimize the non-contact
region caused by body curvature in any posture; (4) our system was covered for durability and easy
set-up; (5) the shielding electrode connected to ground was located at the back of two sensing
electrodes for external noise reduction; (6) the distance between preamplifier and sensing electrode
was shortened to increase signal-to-noise ratio (SNR).
We analyzed the ECG morphology, heart rate (HR) coverage, HRV time and spectral parameters to
evaluate the performance of the system. The increased RR intervals by motion artifacts were filled by
interpolation. Results of experiments show that the non-contact ECG measurement system is effective
to record ECG without any contact to users with high R-peak coverage, low HRV errors, and clear
Sensors 2015, 15 11298

ECG morphology. In addition, we showed that our developed system could detect movement (wake)
by increased RR intervals, and also estimate REM sleep using HRV parameters.

2. Capacitive ECG Measurement System

We developed the capacitively coupled ECG monitoring system as shown in Figure 1, consisting of
sensors for ECG measurement, foam for comfort, and a cover for prevention of pollution.

Figure 1. Configuration of capacitively coupled textile system: (a) size and array of each
electrode in side view; (b) actual design in top view; (c) system cover; (d) experiment
installation during PSG; (e) preamplifier module; and (f) folded system.

2.1. Conductive Fabric

Conductive fabric (WR-260-PCN, A-Jin Electron, Busan, Korea) was chosen considering its
conductivity, durability, weight, elasticity, and sensor suitability, and the entire surface was coated
with nickel and copper. The fabric was a halogen-free product that is harmless to humans. The
thickness was 0.1 mm, and it had excellent electric conductivity (the sheet resistance < 0.05 Ω/sq). The
conductive fabric was used for the sensing electrodes, ground electrodes, and shielding electrode,
shown in Figure 1a,b.

2.2. Polyurethane Foam

Polyurethane foam was used to provide comfort and to reflect the curvature of the body, especially
the waist in a supine posture or the sides in lateral postures. The form had large flexibility and recovery
properties, and also able to reduce the air gaps and non-contact region between the textile electrodes
and the body. The thickness of the foam was 10 mm, and its density was chosen as 35 kg/m3,
considering weight and durability. The foam was attached at the back of conductive textile as shown in
Figure 1a,b.
Sensors 2015, 15 11299

2.3. System Cover

To avoid contamination of the electrodes by external exposure, a system cover was produced using
a 100% cotton fabric, which is general material for bedding. As shown in Figure 1c, five lines on the
cover indicate which side contacts with the body and showed where the upper body should lie on,
taking into account the different heights of the subjects. For example, upper three lines marked the
location of sensing electrodes. Moreover, the backside of the cover had a zipper of length 1.8 m to
separate or install the system.

2.4. Preamplifier

Preamplifiers were used to increase the SNR. Each preamplifier was coated with a silicone
conformal coating (LDC 2577D, Dow Corning Co., Midland, MI, USA) for waterproofing and
protection. Because the thickness of the preamplifier was 3 mm (including on-chips), the preamplifier was
put into a square hole in the foam next to each sensing electrode to reduce the firm feeling (Figure 1e).

2.5. System Design

The system was composed of five conductive textile electrodes, as illustrated in Figure 1a. The front
side had two sensing electrodes for ECG measurement (middle) and two ground electrodes for
common mode noise reduction (upper and lower). The back side had a shielding electrode to reduce
external noise interference, instead of a preamplifier shield made of aluminum plate. Therefore, the
shielding electrode was connected to ground and designed to cover the entire two sensing electrodes in
the back. All conductive textiles were fixed on the polyurethane foam. By attaching additional foam of
thickness 10 mm under each sensing electrode, the area that contacted the body was widened.
Therefore, the sensing electrodes were slightly protruded, but the users were not in any discomfort
while lying on the system, as the foam was very soft.
The total size of the smart mattress was 100 cm × 200 cm, which is equal to the size of a single
mattress. The sizes of the textile electrodes are described in detail in Figure 1a. The size of the upper
ground electrode was chosen considering the general pillow size. The configuration of the sensing
electrodes was designed as a long strip in order to cover various sleep postures and any location in a
bed (17.5 cm × 85 cm). The lower ground electrode was placed under the lower half of the body, and
its size was designed to be as large as possible to get stable signals, because the impedance between
the body and the electrode was decreased by increasing the contact area (76 cm × 85 cm). Eyelets were
used for connecting between the textiles and preamplifier instead of soldering. We also used
conductive textiles to protect the wires from external noise interference and to fix them on the foam.
The actual installation of the system on a bed is shown in Figure 1d, in which the wires were for PSG.
Moreover, the capacitively coupled textile system can be fold for convenient storage like a blanket
(Kolon, Seoul, Korea) (Figure 1f).

2.6. Hardware Specification

A high input impedance amplifier is required to measure ECG through clothes. Figure 2 describes
the electrical circuit of the electrode. The gain of the circuit is as below:
Sensors 2015, 15 11300

Vo Z B // Z A
G s ( s) = = (1)
V s Z C + Z B // Z A

where ZA is the input impedance of the operational amplifier (RA//CA), ZB is the impedance of the
parallel combination of CB and RB, and ZC is the impedance of the cloth and the cover in this system
(RC//CC). Since ZA is sufficiently larger than ZB, ZA can be disregarded. Therefore, the gain is
defined as:
R B + sC C R B RC
G s (s) = (2)
( R B + R C ) + s (C B + C C ) R B R C

In our system, an operational amplifier, OPA124 (Texas Instruments, Dallas, TX, USA), which had
a high input impedance of 1013 Ω || 1 pF was chosen for use as the preamplifier. The hardware module
was composed of an amplifier and three FIR Sallen–Key Butterworth filters: a second-order high-pass
filter (cutoff frequency, fc = 0.5 Hz), a notch filter (UAF42, Texas Instruments, fc = 60 Hz), and an
eight-order low-pass filter (fc = 35 Hz). The total gain was 100 V/V. The overall schematic diagram of
the hardware is illustrated in Figure 3.

Figure 2. Diagram of the electrical circuit of the electrode.

Figure 3. Schematic diagram of the analogue-signal processing module and data acquisition
system (fs: sampling frequency).
Sensors 2015, 15 11301

3. Performance Evaluation

3.1. ECG Morphology

Before performing the PSG experiment, we verified the signal quality of the developed system by
comparing ECG morphology measured from our system to that from the reference system. Seven
healthy male subjects (between 23 and 31 years) participated in the experiment. After measuring the
ECG for supine posture on the capacitive textile system, the ECG was measured again using reference
Ag/AgCl electrodes (Biopac MP 150, Biopac Inc., Goleta, CA, USA) attached on the subject’s back at
the similar location in contact with the textile sensing electrodes when the subject lay on our system.
The correlations of ECG morphologies measured from the proposed system and reference system
were evaluated.

3.2. R-Peak Coverages and HRV Parameters during Sleep

3.2.1. Participants

Nine healthy male subjects (between 23 and 31 years) participated in the experiment. Two subjects
performed daytime PSG and other seven did nighttime PSG. All subjects gave informed consent to
participate, and none of them suffered from a cardiovascular disease or any medical condition.

3.2.2. Protocols

The study was approved by the Institutional Review Board of Seoul National University Hospital
(SNUH), and the developed system was installed on a bed in the Sleep Center at SNUH. The subjects
slept on the system wearing cotton pyjamas. ECG signals from our system were recorded
simultaneously with PSG data using a data acquisition board, NI-DAQ 6221 (National Instruments,
Austin, TX, USA), with a 250 Hz sampling rate. The PSG data contained ECG from Lead II (Ag/AgCl
electrodes), electroencephalogram (EEG), electrooculogram (EOG), electromyogram (EMG) from the
chin, nasal-oral airflow, snoring, and pulse oximetry (SpO2) signals. The PSG data were scored by a
polysomnographic technologist and a sleep physician according to the criteria of Rechtschaffen and
Kales [32].

3.2.3. Data Analysis

The performance of the proposed system was evaluated by comparing the R-peak coverages from
the system with those from Ag/AgCl electrodes (Lead II ECG) during PSG. R-peak coverage was
defined as a percentage of the number of R-peak from capacitive textile system of that from Ag/AgCl
electrodes on PSG system.
We used an internally developed automatic peak detection algorithm and then checked the detected
peaks manually. Moreover, the HRV time and spectral parameters, which are measures of variations in
the HR calculated from the RR intervals in the ECG, were assessed using the normalized error (NE)
between ECG from Lead II and our system [33,34]. NE is defined as follows:
Sensors 2015, 15 11302

X Lead II ECG − X our system

NE = × 100(%) (3)
95% confidence interval length of X Lead II ECG

where X is an HRV parameter in time or frequency domain.

The HRV parameters were calculated in 5 min (10 epochs) window with 90% overlap. Since
motion artifacts lead to missing R-peak, the increased RR intervals (longer than mean + 3 × SD, where
SD is standard deviation) were automatically filled using the Piecewise Cubic Hermite Interpolating
Polynomial method. All signal processing was done with MATLAB 2013b.

4. Results

4.1. ECG Morphology

The ECG morphologies measured from our developed system and reference Ag/AgCl electrodes
were shown in Figure 4. The ECG morphology was filtered with 0.5 Hz–35 Hz. Since the ECGs from
both systems were recorded separately, the ECG waveforms in 5 min term had to be averaged to be
compared. The QRS complex from Ag/AgCl electrodes (red line) was almost same with that from
capacitive textile electrodes (blue line). The mean correlation of ECG morphologies measured from
two systems was 0.96.

Figure 4. The normalized ECG waveforms measured from Ag/AgCl electrodes (red line)
and capacitive textile electrodes (blue line).

4.2. R-Peak Coverages and HRV Parameters during Sleep

Figure 5 illustrates the filtered ECG signals (0.5 Hz–35 Hz) with detected R peaks and RR intervals
from the capacitively coupled textile electrodes and Ag/AgCl electrodes (Lead II ECG) during PSG.
Although the ECG waveforms were different in both systems because the body parts in contact with
the electrodes of each system were different, the developed system reflected the heart beats well even
when the RR intervals in a certain period were shorter or longer than normal.
Sensors 2015, 15 11303

Figure 5. R peaks and RR intervals including (a) shorter RR intervals and (b) longer RR
intervals recorded from Ag/AgCl electrodes (PSG) and capacitive textile electrodes.

ECG signals with motion artifacts are shown in Figure 6. Large and continuous motion artifacts due
to changing sleep posture distorted the ECG signals even in the direct-contact electrodes. However, the
ECG signals recorded from the Ag/AgCl electrodes recovered quickly to get R peaks, whereas the
ECG data from the capacitive textile electrodes took a longer period before obtaining R peaks in
Figure 6a. Small limb movements caused short duration motion artifacts only in the capacitive
measurement system, as shown in Figure 6b.

Figure 6. ECG waveforms with motion artifacts due to (a) sleep posture change and
(b) limb movement, from Ag/AgCl electrodes (PSG) and capacitive textile electrodes.

The total recording time, R-peak coverage, and NE of HRV parameters between Ag/AgCl electrode
system and capacitive textile system are summarized in Tables 1 and 2 for the time and frequency
domains, respectively. The R peaks coverage ranged from 95.5% to 99.7% of the overall night, with a
mean coverage of 98.0%. The smallest mean NE was 0.18% for the mean HR in time domain and
0.15% for the VLF (very low frequency) power in frequency domain. The mean NE of the HRV
parameters was less than 5%, with a range of 0.15%–4.20%.
Sensors 2015, 15 11304

Table 1. R-peak coverage and normalized errors of HRV parameters in time domain
between Ag/AgCl electrodes and capacitive textile electrodes during PSG for nine subjects.
Total Time R-Peak Normalized Errors (%) in Time Domain
Subject a
(h.min.s) Coverage (%) SDNN RMSSD b NN50 c pNN50 d Mean HR e
A 8.16.09 97.97 1.80 1.80 9.23 9.26 0.08
B 8.34.42 99.66 0.45 2.10 0.48 0.48 0.02
C 3.51.24 98.04 1.02 6.72 2.40 2.36 0.50
D 3.23.47 99.35 0.21 3.78 3.01 3.28 0.05
E 8.04.21 99.19 2.26 7.78 0.88 0.95 0.03
F 8.21.20 97.41 2.09 5.74 4.58 4.45 0.09
G 9.58.45 96.87 0.75 3.18 4.53 4.90 0.11
H 8.43.54 95.47 3.54 2.20 1.54 1.77 0.59
I 7.56.58 98.06 1.81 4.47 2.45 2.51 0.16
Mean 98.00 1.55 4.20 3.23 3.33 0.18
a
SDNN: Standard deviation of the RR intervals; b RMSSD: Root mean square successive differences of RR
intervals; c NN50: Number of pairs of successive RR intervals that differ by more than 50 ms; d pNN50:
Proportion of NN50 from the total number of RR intervals; e Mean HR: Mean of heart rate.

Table 2. R-peak coverage and normalized errors of HRV parameters in frequency domain
between Ag/AgCl electrodes and capacitive textile electrodes during PSG for nine subjects.
Total Time R-Peak Normalized Errors (%) in Frequency Domain
Subject a
(h.min.s) Coverage (%) LF HF b LF/HF c nLF d nHF e VLF f nVLF g
A 8.16.09 97.97 0.20 0.26 0.70 0.74 0.74 0.16 1.04
B 8.34.42 99.66 0.16 0.38 1.65 1.44 1.44 0.01 0.18
C 3.51.24 98.04 1.66 1.38 1.29 1.37 1.37 0.39 5.23
D 3.23.47 99.35 0.15 0.48 2.00 1.87 1.88 0.01 0.49
E 8.04.21 99.19 0.08 5.84 13.09 15.0 15.1 0.10 5.39
F 8.21.20 97.41 0.35 1.08 2.04 2.13 2.13 0.18 1.71
G 9.58.45 96.87 0.06 0.20 1.05 1.33 1.34 0.01 0.11
H 8.43.54 95.47 0.27 1.16 3.56 2.83 2.84 0.25 0.55
I 7.56.58 98.06 0.50 0.95 1.34 1.24 1.24 0.24 5.41
Mean 98.00 0.38 1.30 2.97 3.11 3.12 0.15 2.23
a b
LF: Power in low frequency range (0.04–0.15 Hz).; HF: Power in high frequency range (0.15–0.4 Hz);
c
LF/HF: Ratio of LF to HF; d nLF: LF power in normalized units; e nHF: HF power in normalized units;
f
VLF: Power in very low frequency range (0.0033–0.04 Hz); g nVLF: VLF power in normalized units.

5. Discussion

The width of sensing electrodes was designed to be large enough to cover a single mattress, because
it is important to record ECG regarding any position on a bed during sleep. However, the increased
surface area of sensing electrodes also increased the area where the body does not contact with, and
this causes more noise that affect the quality of ECG signals. In addition, Wu and Zhang [18] reported
that the amplitudes of QRS complexes acquired from the hands placed inside the electrodes were
smaller than from those placed out of the electrode region. To handle these problems, the shielding
electrode connected to ground was located at the back of the sensing electrodes, and the distance
between the sensing electrode and preamplifier was adjusted as short as possible. Moreover, all the
wires from the sensing electrodes and preamplifier were shielded as well to increase the SNR.
Sensors 2015, 15 11305

Therefore, our system was able to record ECG signals with good quality for all subjects no matter what
kind of postures and positions they had during sleep.
As shown in Figure 2, the body weight changes the CB in the electrode, because as the body weight
increases, the compression on the form between the shielding electrode and the sensing electrode also
increases. Despite the fact that the body weight could affect the gain of the system, the reason for using
a very soft form instead of a hard and rigid material was to provide comfort and to reflect the curvature
of the body for more contact region during sleep. In our experiment, the weights of subjects were
ranged from 57 to 85 kg. Making a stable contact environment was more important to measure ECG
for long term monitoring because we used the filtered ECG signals.
The ground of the reference system was same with that of our capacitive system, because ECGs
were acquired through NIDAQ. Therefore, the ground of PSG system could decrease the noise of the
capacitive measurement. However, there was no problem to analyze ECG signals with the increased
baseline noise since filtering was used before detecting R-peak. Unlike some other researches on
capacitively coupled ECG measurement [19,35–37], this study did not use a driven-right-leg circuit,
instead of ground, for common mode noise reduction. Since the driven signals give feedback the
inverted common-mode voltage to the subject’s body, the signals can affect the EEG measurement
from PSG system. Therefore, we used the ground system so that PSG could be measured
simultaneously. However, there would be no issue in using a driven-right-leg circuit in our system,
which would provide even better performances.
The percentage of R-peak coverage was proportional to the ratio of wake duration to total time
(defined as wake ratio) during PSG. For example, subject B had a 2.4% wake ratio during the total
time of 8.34 h, and the percentage of R-peak coverage for subject B was 99.7%. On the other hand,
subject H had a 27.6% of wake ratio during 8.43 h, and the percentage of R-peak coverage was 95.5%.
Figure 7 shows the sleep stages of subject H and his movements (red circles) detected by increased RR
intervals (longer than mean + 3 × SD) from the textile electrodes. It can be observed that this subject
did not sleep well, because of the unfamiliar environment and many wire sensors. He even sat on the
bed for 10 min, tossed and turned a lot during PSG, and then tried to go back to sleep, resulting in a lot
of movement in the beginning stage of sleep. Nevertheless, the NEs of HRV measures from subject H
were less than 4%, which is under average.

Figure 7. Sleep stages of subject H during PSG and movement (red circles) detected by
increased RR intervals measured from the capacitive textile system.
Sensors 2015, 15 11306

A direct comparison between the performance of our system and the results reported by other
studies is difficult because of the use of different datasets, measurement technology, and window size
for analyses. Nevertheless, the HR coverage and root mean square errors (RMSEs) of HR are
summarized in Table 3 for various technologies and electrode designs.

Table 3. Performance summary of related works depending on technologies and electrode design.
Performances
Reference Technologies Used Electrode Design
Parameters Results
Ishijima [13] Textile, contact Pillow, foot mat HR coverage 83%–93%
Devot et al. [14] Textile, contact Pillow, foot mat HR coverage Average 81.89%
Bed, array
Park et al. [15] Textile, contact Usable ECG data Average 84.2%
(multi-channel)
Average 94.9%
HR coverage
(85.05%–98.98%)
Bed, array RMSE a for RR interval 4.48 ms
Peltokangas et al. [16] Textile, contact
(multi-channel) RMSE of HR 0.27 bpm
b
Relative MAE of
0.89%, 3.90%, 2.20%
LF, HF, LF/HF
RR interval coverage 98%
Wu and Zhang [18] Textile, capacitive Bed, strips RMSE for RR interval 18.3 ms
RMSE of HR 1.24 bpm
HR coverage Average 98%
RMSE for RR interval 1.36 ms
RMSE of HR 0.09 bpm
Our system Textile, capacitive Bed, strips Relative MAE of
0.80%, 0.05%
SDNN, mean HR
Relative MAE of
0.29%, 0.93%, 0.88%
LF, HF, LF/HF
a
RMSE: Root mean square error; b MAE: Mean absolute error.

To compare directly with the existing methods in literature, we also calculated the RMSEs and
relative mean absolute error (MAE) between the Ag/AgCl system (PSG) and our developed system.
Our system had a high average of HR coverage and a remarkably low RMSEs for RR interval and HR.
For HRV parameters, we could only compare with the results obtained from direct-contact
methods. Moreover, the relative MAE (%) of our system was considerably lower than those of
Peltokangas et al. [16].
Our developed system, which is based on a capacitively coupled measurement method, is sensitive
to motion artifacts, as shown in Figure 6. The failed R-peak detection usually occurred during sleep
posture change or small limb movements. Therefore, since movement usually takes place during wake
time, our system has the potential to provide additional information about wake/sleep detection, as in
Figure 7 [38].
As our system uses the capacitively coupled method, there are some limitations. First, the
material of clothes affects the RC and CC of the electrode that are related to gain as shown in Figure 2.
Lee et al. concluded that the gain of cotton was higher than that of polyester when using same bias
Sensors 2015, 15 11307

resistor [39]. Second, thick clothes or the blanket should not be caught between the body and the
cover. This can happen when changing postures, especially from lateral posture to supine posture.
Although some non-electrical methods that do not have this problem have been studied for monitoring
heart function, the mechanical signals are more vulnerable to movement and more difficult for beats
detection, especially ectopic beats that do not affect the pumping mechanism of the heart [40].
Therefore, unconstrained and contactless ECG recording is the best approach for long-term
healthcare monitoring.
As shown in Tables 1 and 2, the HRV parameters from our system in the time and frequency
domains are guaranteed with low errors, especially in mean HR, LF power, HF power, and VLF
power. These are useful clinical measures for detecting hypertension, sudden death or cardiac arrest,
myocardial infarction, and diabetes [25]. Moreover, the parameters can also provide information on the
stress state of the body as the HRV spectral measures are related to changes in the sympathetic and
parasympathetic activity of the autonomic nervous system. In addition, the ECG data from our system
can be utilized in sleep analyses through the HRV analyses. Figure 8 shows the sleep stages, HF
power, and LF/HF ratio from the capacitively measured ECG of subject B during PSG.

Figure 8. Sleep stages: (a) HF power and (b) LF/HF ratio from capacitive textile system of
subject B during PSG.

These parameters follow the same trend, with HF power decreasing and LF/HF ratio increasing
during REM sleep, as reported by sleep-related literatures [14,26]. Therefore, REM and non-REM
sleep can be classified using the changes in the spectral powers from our system. Furthermore, as
shown in Figure 5, the quality of the ECG signals provided by our system can be adequate for the
detection of abnormal states during sleep such as arrhythmia and ectopic beats.

6. Conclusions

In this study, we have developed a non-contact ECG measurement system on a bed using
conductive textiles for long-term monitoring of heart or cardiovascular function. Our approach was
based on designing a more user-friendly system by configuring the array and size of electrodes to
measure ECG with no constraints of posture and position in a bed, using a foam layer to provide
Sensors 2015, 15 11308

comfort and follow the curvature of the body, and using a system cover for durability and easy
installation. Moreover, we tested the system through HR and HRV analyses to see its potential for
monitoring the sleep stages, which is important for observing the physical and psychological states.
Our results suggest that the capacitively coupled textile system is suitable for daily healthcare
monitoring, with high R-peak coverage, low HRV errors, clear ECG morphology, and accurate R
peaks when compared with a commercial system.

Acknowledgments

We would like to thank Kolon, Bionet, and the Sleep Center at Seoul National University Hospital
in the Republic of Korea. This study was supported by the National Research Foundation
of Korea (NRF) grant funded by the Ministry of Science, ICT & Future Planning
(No. NRF-2012R1A2A2A02010714), in part by BK21 Plus Program through the National Research
Foundation of Korea (NRF) funded by the Ministry of Education (grant Nos. 22A20130011025).

Author Contributions

Hong Ji Lee and Su Hwan Hwang conceived, designed, and performed the experiments;
Hong Ji Lee analyzed the data; Su Hwan Hwang, Hee Nam Yoon, and Won Kyu Lee contributed
analysis tools; Hong Ji Lee and Kwang Suk Park wrote and reviewed the manuscript.

Conflicts of Interest

The authors declare no conflict of interest.

References

1. Lim, Y.G.; Kim, K.K.; Park, K.S. ECG measurement on a chair without conductive contact.
IEEE Trans. Biomed. Eng. 2006, 53, 956–959.
2. Tamura, T.; Togawa, T.; Ogawa, M.; Yoda, M. Fully automated health monitoring system in the
home. Med. Eng. Phys. 1998, 20, 573–579.
3. Wiard, R.M.; Inan, O.T.; Argyres, B.; Etemadi, M.; Kovacs, G.T.; Giovangrandi, L. Automatic
detection of motion artifacts in the ballistocardiogram measured on a modified bathroom scale.
Med. Biol. Eng. Comput. 2011, 49, 213–220.
4. Seo, J.; Choi, J.; Choi, B.; Jeong, D.-U.; Park, K. The development of a nonintrusive home-based
physiologic signal measurement system. Telemed. J. e-Health 2005, 11, 487–495.
5. Lim, Y.G.; Hong, K.H.; Kim, K.K.; Shin, J.H.; Lee, S.M.; Chung, G.S.; Baek, H.J.; Jeong, D.-U.;
Park, K.S. Monitoring physiological signals using nonintrusive sensors installed in daily life
equipment. Biomed. Eng. Lett. 2011, 1, 11–20.
6. Hwang, S.H.; Lee, H.J.; Yoon, H.N.; Jung, D.W.; Lee, Y.G.; Lee, Y.J.; Jeong, D.U.; Park, K.S.
Unconstrained sleep apnea monitoring using Polyvinylidene Fluoride Film-Based Sensor.
IEEE Trans. Biomed. Eng. 2014, 61, 2125–2134.
Sensors 2015, 15 11309

7. Wong, M.; Leung H.; Pickwell-MacPherson, E.; Gu, W.; Zhang, Y. Contactless recording of
photoplethysmogram on a sleeping bed. In Proceedings of the 31st Annual International
Conference of the IEEE Engineering in Medicine and Biology Society(EMBS), Minneapolis, MN,
USA, 3–9 September 2009; pp. 907–910.
8. Tanaka, S.; Matsumoto, Y.; Wakimoto, K. Unconstrained and non-invasive measurement of
heart-beat and respiration periods using a phonocardiographic sensor. Med. Biol. Eng. Comput.
2002, 40, 246–252.
9. Watanabe, K.; Watanabe, T.; Watanabe, H.; Ando, H.; Ishikawa, T.; Kobayashi, K. Noninvasive
measurement of heartbeat, respiration, snoring and body movements of a subject in bed via a
pneumatic method. IEEE Trans. Biomed. Eng. 2005, 52, 2100–2107.
10. Mack, D.C.; Patrie, J.T.; Suratt, P.M.; Felder, R.A.; Alwan, M.A. Development and preliminary
validation of heart rate and breathing rate detection using a passive, ballistocardiography-based
sleep monitoring system. IEEE Trans. Inf. Technol. Biomed. 2009, 13, 111–120.
11. Junnila, S.; Akhbardeh, A.; Barna, L.C.; Defee, I.; Varri, A. A wireless ballistocardiographic
chair. In Proceedings of the 28th Annual International Conference of the IEEE Engineering in
Medicine and Biology Society(EMBS), New York, NY, USA, 30 August–3 September 2006;
pp. 5932–5935.
12. Kim, J.; Park, J.; Kim, K.; Chee, Y.; Lim, Y.; Park, K. Development of a nonintrusive blood
pressure estimation system for computer users. Telemed. J. e-Health 2007, 13, 57–64.
13. Ishijima, M. Cardiopulmonary monitoring by textile electrodes without subject-awareness of
being monitored. Med. Biol. Eng. Comput. 1997, 35, 685–690.
14. Devot, S.; Bianchi, A.; Naujokat, E.; Mendez, M.; Brauers, A.; Cerutti, S. Sleep monitoring
through a textile recording system. In Proceedings of the 29th Annual International Conference of
the IEEE Engineering in Medicine and Biology Society (EMBS), Lyon, France, 22–26 August
2007; pp. 2560–2563.
15. Park, S.B.; Noh, Y.S.; Park, S.J.; Yoon, H.R. An improved algorithm for respiration signal
extraction from electrocardiogram measured by conductive textile electrodes using instantaneous
frequency estimation. Med. Biol. Eng. Comput. 2008, 46, 147–158.
16. Peltokangas, M.; Verho, J.; Vehkaoja, A. Night-time EKG and HRV monitoring with bed sheet
integrated textile electrodes. IEEE Trans. Inf. Technol. Biomed. 2012, 16, 935–942.
17. Lim, Y.G.; Kim, K.K.; Park, K.S. ECG recording on a bed during sleep without direct
skin-contact. IEEE Trans. Biomed. Eng. 2007, 54, 718–725.
18. Wu, K.; Zhang, Y. Contactless and continuous monitoring of heart electric activities through
clothes on a sleeping bed. In Proceedings of the International Conference on Information
Technology and Applications in Biomedicine(ITAB), Shenzhen, China, 30–31 May 2008;
pp. 282–285.
19. Ueno, A.; Yama, Y. Unconstrained monitoring of ECG and respiratory variation in infants with
underwear during sleep using a bed-sheet electrode unit. In Proceedings of the 30th Annual
International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS),
Vancouver, BC, USA, 20–25 August 2008; pp. 2329–2332.
Sensors 2015, 15 11310

20. Ishida, S.; Shiozawa, N.; Fujiwara, Y.; Makikawa, M. Electrocardiogram Measurement during
Sleep with Wearing Clothes Using Capacitively-Coupled Electrodes. In Proceedings of the 29th
Annual International Conference of the IEEE Engineering in Medicine and Biology Society
(EMBS), Lyon, France, 22–26 August 2007; pp. 2647–2650.
21. Chi, Y.M.; Jung, T.-P.; Cauwenberghs, G. Dry-contact and noncontact biopotential electrodes:
Methodological review. IEEE Rev. Biomed. Eng. 2010, 3, 106–119.
22. Griffith, M.E.; Portnoy, W.M.; Stotts, L.J. Improved capacitive electrocardiogram electrodes for
burn applications. Med. Biol. Eng. Comput. 1979, 17, 641–646.
23. Searle, A.; Kirkup, L. A direct comparison of wet, dry and insulating bioelectric recording
electrodes. Physiol. Meas. 2000, 21, 271–283.
24. Ueno, A.; Akabane, Y.; Kato, T.; Hoshino, H.; Kataoka, S.; Ishiyama, Y. Capacitive Sensing of
Electrocardiographic Potential Through Cloth From the Dorsal Surface of the Body in a Supine
Position: A Preliminary Study. IEEE Trans. Biomed. Eng. 2007, 54, 759–766.
25. Malik, M.; Bigger, J.T.; Camm, A.J.; Kleiger, R.E.; Malliani, A.; Moss, A.J.; Schwartz, P.J. Heart
rate variability: standards of measurement, physiological interpretation, and clinical use.
Eur. Heart J. 1996, 17, 354–381.
26. Kesper, K.; Canisius, S.; Penzel, T.; Ploch, T.; Cassel, W. ECG signal analysis for the assessment
of sleep-disordered breathing and sleep pattern. Med. Biol. Eng. Comput. 2012, 50, 135–144.
27. De Bacquer, D.; de Backer, G.; Kornitzer, M.; Blackburn, H. Prognostic value of ECG findings
for total, cardiovascular disease, and coronary heart disease death in men and women. Heart 1998,
80, 570–577.
28. Hall, M.; Vasko, R.; Buysse, D.; Ombao, H.; Chen, Q.; Cashmere, J.D.; Kupfer, D.; Thayer, J.F.
Acute stress affects heart rate variability during sleep. Psychosomat. Med. 2004, 66, 56–62.
29. Irwin, M. Effects of Sleep and Sleep Loss on Immunity and Cytokines. Brain Behav. Immun.
2002, 16, 503–512.
30. Benz, R.L.; Pressman, M.R.; Hovick, E.T.; Peterson, D.D. A preliminary study of the effects of
correction of anemia with recombinant human erythropoietin therapy on sleep, sleep disorders,
and daytime sleepiness in hemodialysis patients (The SLEEPO study). Am. J. Kidney Dis. 1999,
34, 1089–1095.
31. Schenck, C.; Mahowald, M. Motor dyscontrol in narcolepsy: Rapid-eye-movement (REM) sleep
without atonia and REM sleep behavior disorder. Ann. Neurol. 1992, 32, 3–10.
32. Rechtschaffen, A.; Kales, A. A Manual of Standardized Terminology, Techniques and Scoring
System for Sleep Stages of Human Subjects; Public Health Service, U.S. Government Printing
Office: Washington, DC, USA, 1968.
33. Limpert, E.; Stahel, W.A.; Abbt, M. Log-normal distributions across the sciences: Keys and clues.
BioScience 2001, 51, 341–352.
34. Kim, K.K.; Kim, J.S.; Lim, Y.G.; Park, K.S. The effect of missing RR-interval data on heart rate
variability analysis in the frequency domain. Physiol. Meas. 2009, 30, 1039–1050.
35. Spinelli, E.M.; Martinez, N.H.; Mayosky, M.A. A transconductance driven right-leg circuit.
IEEE Trans. Biomed. Eng. 1999, 46, 1466–1470.
Sensors 2015, 15 11311

36. Kim, K.K.; Lim, Y.K.; Park, K.S. Common mode noise cancellation for electrically non-contact
ECG measurement system on a chair. In Proceedings of the 27th Annual International Conference
of the IEEE Engineering in Medicine and Biology Society (EMBS), Shanghai, China, 17–18
January 2006; pp. 5881–5883.
37. Nemati, E.; Deen, M.J.; Mondal, T. A wireless wearable ECG sensor for long-term applications.
IEEE Commun. Mag. 2012, 50, 36–43.
38. Jung, D.W.; Hwang, S.H.; Yoon, H.N.; Lee, Y.J.G.; Jeong, D.U.; Park, K.S. Nocturnal
Awakening and Sleep Efficiency Estimation Using Unobtrusively Measured Ballistocardiogram.
IEEE Trans. Biomed. Eng. 2014, 61, 131–138.
39. Lee, J.S.; Heo, J.; Lee, W.K.; Lim, Y.K.; Kim, Y.H.; Park, K.S. Flexible capacitive electrodes for
minimizing motion artifacts in ambulatory electrocardiograms. Sensors 2014, 14, 14732–14743.
40. Kortelainen, J.; Mendez, M.; Bianchi, A.; Matteucci, M.; Cerutti, S. Sleep staging based on
signals acquired through bed sensor. IEEE Trans. Inf. Technol. Biomed. 2010, 14, 776–785.

© 2015 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article
distributed under the terms and conditions of the Creative Commons Attribution license
(http://creativecommons.org/licenses/by/4.0/).