SIGNAL QUALITY ASSESSMENT FOR CAPACITIVE ECG

MONITORING SYSTEMS USING BODY-SENSOR-IMPEDANCE

Stephan Heuer, Sebastian Chiriac, Malte Kirst

FZI Research Center for Information Technology, Haid-und-Neu-Str. 10-14, 76131 Karlsruhe, Germany

Adnene Gharbi, Wilhelm Stork

Institute for Information Processing Technology, Karlsruhe Institute of Technology, Karlsruhe, Germany

Keywords:

Capacitive ECG measurement, Unobtrusive ECG measurement, Non-contact ECG, Active electrodes, Artifact

detection, Signal quality, Context signals, QRS detection, Adaptive ﬁltering.

Abstract:

Contactless capacitive ECG measurement is an unobtrusive way of acquiring cardiovascular data. However,

movement artifacts present a common problem with this technique. A means of assessing signal quality and

conﬁdence is therefore desirable. In this paper we present a capacitive ECG measurement system with an

integrated module that constantly monitors the electrode-body-impedance. Moreover, we present a method

to derive an artifact level signal from this electrode-body-impedance that can be used to estimate the signal

quality of the capacitive ECG measurement. First results of measurements with this system are shown.

1 INTRODUCTION

Continuous cardiovascular health monitoring systems

enable a wide range of applications both in the do-

main of physiological long-term monitoring and in

psychophysiological monitoring. Both ﬁelds of appli-

cation require unobtrusive system concepts, that are

for the most part realized as wearable devices and

smart clothing or as ambient unobtrusive sensor sys-

tems. Examples for such systems can be found in

(Park et al., 2006; Lamparth et al., 2009; Lim et al.,

2006).

The amount of data acquired with said systems is

often large, especially in long-term applications, de-

manding automatic (pre-)analysis. Reliable accurate

analysis is yet difﬁcult to achieve due to artifacts re-

sulting from motion during daily routine activities.

Additionally, there is often a trade-off between sig-

nal quality and sensor integration aspects. Taking this

into consideration, system design must not only con-

sider sensor development for the respective applica-

tion but also requires tight integration of well adapted

algorithms for automatic signal analysis.

In ECG monitoring tasks, artifacts from exter-

nal noise and motion are the most common fac-

tors that impair signal quality. Reliable detection of

R-peaks, for example as a necessary step in heart

rhythm analysis, is not given during intervals with

artifacts, leading to wrong detection results. This

is especially a problem with contactless, capacitively

coupled ECG systems, speciﬁcally when coupling is

weak and body-sensor-distance is not constant, as for

example in chair-integrated solutions (Aleksandrow-

icz et al., 2007). It is therefore desirable to improve

the detection accuracy in automatic ECG analysis for

contactless ECG measurement systems.

2 RELATED WORK

Using a mobile ECG measurement system with gal-

vanic dry electrodes, (Ottenbacher et al., 2008) simul-

taneously acquire ECG and electrode-skin-impedance

data (as well as acceleration data). In order to validate

the results a reference ECG signal with wet electrodes

is recorded in parallel.

They propose a method to improve an automatic

QRS detector by calculating an artifact level from the

recorded electrode-skin-impedance using adaptive ﬁl-

tering and post-processing. Thus they are able to

mark artifact regions in the ECG signal and exclude

them from automatic detection, considerably decreas-

ing false positive and false negative QRS detections.

Adapting parts of this method to a contactless

454

Heuer S., Chiriac S., Kirst M., Gharbi A. and Stork W..

SIGNAL QUALITY ASSESSMENT FOR CAPACITIVE ECG MONITORING SYSTEMS USING BODY-SENSOR-IMPEDANCE.

DOI: 10.5220/0003160004540458

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2011), pages 454-458

ISBN: 978-989-8425-35-5

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Block diagram of the developed contactless ECG

system with integrated capacitance measurement module.

The textile electrode C

carries the excitation signal gen-

erated by the AD7152. Body movements generate capaci-

tance variations that are sensed by C

and C

ECG-system in an air-plane seat, (Schumm et al.,

2009) propose a method to predict the signal quality

of a contactless ECG recording. A number of sta-

tistical measures derived from the ECG signal itself

as well as additional pressure sensors attached to the

back of the capacitive electrodes are used as features

to derive a quality signal for the measurement.

3 METHOD

The most prominent source of artifacts in capacitive

ECG measurement is the relative movement between

the body and the capacitive electrodes. This changes

the capacitive coupling between sensor and body, re-

sulting in a voltage peak in the signal if a voltage

difference between the subject and the measurement

system exists. Additionally, CMRR of the differen-

tial stage is reduced, because a mismatch in source

impedances occurs.

As impedance between sensor and body changes

with movement, we propose a method to monitor this

impedance and use it as an indicator for the quality of

the sensor-body-contact. We therefore present an un-

obtrusive measurement system with capacitive ECG

electrodes that records the sensor-body-capacitance

of each electrode in addition to a differential capa-

citive ECG lead.

With these signals we derive an artifact indicator

signal similar to (Ottenbacher et al., 2008) that can

be used in conjunction with automatic QRS detec-

tion algorithms to reduce false detections and improve

the overall detection quality in unobtrusive ECG mea-

surement applications.

4 INSTRUMENTATION

4.1 Measurement System

For the experiment a contactless capacitive ECG

measurement system has been developed and inte-

grated into the backrest of a chair. Additionally, we

have integrated a capacitance measurement module

in order to continuously monitor the CCE-to-body

capacitance.

The overall system structure is shown in Figure 1

and consists of

• a battery-powered, wireless 16-bit data acquisi-

tion platform with an analog front-end for capa-

citive coupled ECG electrodes,

• a capacitive driven-seat electrode (as described by

(Keun Kim et al., 2005)) that can be disabled, de-

pending on common-mode noise level,

• active capacitively coupled ECG electrodes

(CCEs) with a dedicated sensor area for sensor-

to-body capacitance measurement (C

1,2

) and

• a two channel capacitance measurement module

featuring the AD7152 capacitance-to-digital con-

verter by Analog Devices (Analog Devices Inc.,

2008).

The two combined CCE/Capacitance-monitoring

electrodes are realized as multi layer PCBs with iso-

lated sensor areas for ECG and capacitance measure-

ment, details of such an electrode are shown in Fig-

ure 2.

Figure 2: Contactless ECG chair with combined electrodes

for capacitance and ECG measurement. Measurements can

be performed with or without backrest cover.

SIGNAL QUALITY ASSESSMENT FOR CAPACITIVE ECG MONITORING SYSTEMS USING

BODY-SENSOR-IMPEDANCE

455

The inner area of each electrode (A

CCE

≈ 28 cm

)

was connected to the capacitive ECG front-end

whereas the outer ring (A

1,2

≈ 19 cm

) was con-

nected to the input of the capacitance-to-digital mod-

ule (compare Figure 1). The ECG sampling rate was

set to 500 Hz while the capacitance measurement was

running with 200 Hz per channel, the maximum sam-

pling rate possible with the AD7152.

4.2 Capacitance Measurement

Conﬁguration

In order for the AD7152 to perform the capacitance

measurement, a square wave excitation signal with

≈ 32 kHz and V

= 3.2 V is generated on-chip.

The value of a capacitance connected between the ex-

citation signal output and the measurement input of

the chip is directly converted to a 12-bit digital value

and can be read via an I

C-compatible interface. The

device offers two channels that can either be operated

in single-ended mode or in differential mode (Analog

Devices Inc., 2008).

Using one-channel differential mode in our setup,

we have connected the excitation signal to a large

sheet of conductive textile (A

≈ 300 cm

) that was

located on the seat area, as indicated in Figure 2.

As A

 A

1,2

, this setup can be considered a “Hu-

man Transmitter” (Zimmerman et al., 1995). The ex-

citation signal becomes a common-mode signal on

the body and movements of the upper body result

in capacitance changes that can be registered by the

AD7152.

The device’s maximum input range is ±2 pF in

differential mode. Common-mode capacitances of

up to 5 pF can be compensated on-chip. In our ex-

periment, we intended to monitor movements within

a range of ∆x = 0.5 mm. . . 5 mm. This results in a

coupling capacitance range between body and ring-

electrode of ∆C

= 47 pF. . . 4.7 pF when the body-

electrode contact is considered a plate capacitor with

cotton (ε

= 1.4) as dielectric.

In order to adjust the coupling capacitance ∆C

to the

chip’s input range, series capacitors of 4.7 pF were

inserted between the electrodes and the inputs of the

AD7152. This results in an effective capacitance

range of ∆C

e f f

= 1.9 pF per electrode with an offset

of C

= 2.4 pF.

5 ARTIFACT DETECTION

ALGORITHM

The motion artifact detection algorithm we have im-

plemented consists basically of four steps: At ﬁrst an

artifact level representing the intensity of artifacts in

the ECG signal is computed. The generated artifact

level is post-processed and in a third step it is con-

verted by means of a threshold detector into a binary

artifact indicator signal. As a last step, this artifact in-

dicator signal is logically ANDed with an additional

parameter derived directly from the ECG signal wave-

form. This signal is ﬁnally used to mark artifact re-

gions in the original ECG signal.

Artifact Level from Capacitance Data. The dif-

ferential capacitance signal from the AD7152 was

high pass ﬁltered to remove offsets with a 4

order

Butterworth ﬁlter, cutting off signals below 1 Hz, and

the absolute value was taken. Then adaptive ﬁltering

(Haykin, 2001) was applied in order to estimate arti-

facts in the ECG signal from the capacitance signal.

A ﬁlter with LMS algorithm has been used, the ﬁlter

length was set to 0.2 s.

Figure 3: Signal processing chain: the artifact indicator sig-

nal is derived from ECG and impedance data.

Post-processing & Threshold Filtering. As a post-

processing step the signal containing the artifact level

from the previous adaptive ﬁlter step was squared and

ﬁltered with a moving average ﬁlter. The ﬁlter length

was set to 0.5 s.

This was the input to the threshold detection that

transforms the continuous artifact level signal into a

binary artifact indicator signal.

BIOSIGNALS 2011 - International Conference on Bio-inspired Systems and Signal Processing

456

0 . 0 0 5 0 . 0 1 0 . 0 1 5 0 . 0 2

Time [s]

e x

[V]

Figure 4: The 32 kHz excitation signal shows gaps every

7 ms, resulting in 140 Hz common-mode noise in the ECG

signal if no further measures are taken.

Final Artifact Marker Signal. Supplementary to

the artifact indicator, we have taken an additional pa-

rameter into account: using the original ECG data, we

verify that the signal lies within the range of [10, 90]%

LSB

ADC

of the analog-digital-converter, in order to

eliminate regions where the signal is near saturation.

Then the artifact indicator and the in-range indica-

tor were logically ANDed to generate the ﬁnal artifact

marker signal. When this signal is high, the corre-

sponding region in the ECG signal is marked as arti-

fact region. Furthermore, artifact regions that do not

have a minimum distance of 1 s are joined.

6 EVALUATION & RESULTS

With the system presented above, we have conducted

measurements with several subjects. The driven-

seat circuit was not necessary because the systems’

coupling to earth ground was very weak due to the

battery-powered, wireless concept. Yet with the ca-

pacitance measurement circuit enabled, we observed

a large amount of common-mode noise on the ECG

signal with a frequency of approximately 140 Hz, the

source of which was not obvious at ﬁrst. A closer

look at the excitation signal of the AD7152 revealed

that even though the device was used in one-channel

differential mode with continuous conversion, gaps in

the excitation signal occurred with a distance of ap-

proximately 7 ms (see Figure 4). This was visible as

common-mode noise in the ECG signal.

Enabling the driven-seat circuit to remove this

noise was not an option, as a 180

◦

phase shift is neces-

sary to destructively interfere with the undesired fre-

quencies. In the case of 140 Hz, this would result in a

rather high bandwidth of the driven-seat stage. Then,

the excitation signal of 32 kHz would have been at-

tenuated too, resulting in incorrect capacitance mea-

surement values. The solution for us was to simply

narrow the bandwidth of the CCEs by adding 50 Hz

low-pass ﬁlters.

6.1 Artifact Signal

The proposed algorithm generates an artifact signal

with very high dynamics. For slight capacitance

changes generated by breathing and light movements,

the signal amplitudes stay very low. Higher move-

ment amplitudes can increase the artifact amplitudes

up to 4 orders of magnitude. Hence, a threshold could

be identiﬁed that separates regular ECG episodes

from the ones with artifacts. First results showed, that

no adaptation of the threshold was necessary for dif-

ferent subjects.

6.2 Sample Measurement

Figure 5 ﬁnally shows an example of the computa-

tion of artifact regions for a capacitive ECG signal

recorded with the system proposed in this paper. The

gray trace represents the artifact level generated by

adaptive ﬁltering of the differential capacitance values

captured by the AD7152. Above a certain threshold,

the artifact indicator goes high. The artifact region

can be marked in the ECG signal.

7 SUMMARY & OUTLOOK

Non-contact capacitive ECG systems represent an un-

obtrusive way to acquire cardiovascular data. Yet, due

to the measurement principle, these systems are sensi-

tive to motion artifacts. For automatic signal analysis,

a means of estimating signal quality is therefore de-

sired.

Combining a capacitive ECG measurement sys-

tem with a capacitance-to-digital conversion de-

vice, we have presented an unobtrusive ECG sys-

tem that continuously monitors the electrode-body-

capacitance. This capacitance value represents a con-

text signal that can be used to derive a quality signal

for the capacitive ECG measurement.

Using adaptive ﬁltering and post-processing we

were able to show ﬁrst results of the system perfor-

mance. With the system, it is possible to mark re-

gions with artifacts in the capacitive ECG signal, and

exclude them from further processing thus improving

automatic analyzability.

Currently we are building a database with mea-

surements recorded with our system. With the data

we will be able to quantify the improvement of au-

tomatic analysis of unobtrusively acquired ECG data

due to our method. For automatic QRS detection, a

considerable decrease in false (positive and negative)

detections due to the proposed method should be pos-

sible, yielding an increase in sensitivity and positive

SIGNAL QUALITY ASSESSMENT FOR CAPACITIVE ECG MONITORING SYSTEMS USING

BODY-SENSOR-IMPEDANCE

457

10 20 30 40 50 60

−2

−1

Time [s]

Amplitude [a.u.]

Figure 5: Using the adaptively ﬁltered capacitance signal (lower trace) it is possible to generate an indicator signal that can

be used to mark artifact regions (shaded) in a capacitive ECG measurement.

predictivity. Excluding artifact-contaminated parts of

the signal also prevents the internal thresholds of the

QRS-detector from assuming suboptimal values. The

evaluation will also help to understand the applica-

tion limits of the system and the applicability of the

method.

Further work will comprise the improvement of

the artifact detection algorithm by optimizing the

adaptive ﬁlter parameters and timing parameters, as

well as investigations whether the system can also be

used for the correction of artifacts.

REFERENCES

Aleksandrowicz, A., Walter, M., and Leonhardt, S. (2007).

Wireless ECG measurement system with capacitive

coupling. Biomedizinische Technik. Biomedical engi-

neering, 52(2):185–92.

Analog Devices Inc. (2008). AD7152 Datasheet:12-Bit

Capacitance-to-Digital Converter.

Haykin, S. (2001). Adaptive Filter Theory. Prentice Hall,

4th edition.

Keun Kim, K., Kyu Lim, Y., and Suk Park, K. (2005).

Common mode noise cancellation for electrically non-

contact ECG measurement system on a chair. In Engi-

neering in Medicine and Biology Society, Annual In-

ternational Conference of the IEEE, volume 6, pages

5881–3.

Lamparth, S., Fuhrhop, S., Kirst, M., Wagner, G., and Ot-

tenbacher, J. (2009). A Mobile Device for Textile-

integrated Long-term ECG Monitoring. In Dössel, O.

and Schlegel, W. C., editors, World Congress on Med-

ical Physics and Biomedical Engineering, Septem-

ber 7-12, 2009, Munich, Germany, volume 25/5 of

IFMBE Proceedings, pages 278–281, Berlin, Heidel-

berg. Springer Berlin Heidelberg.

Lim, Y. G., Kim, K. K., and Park, K. S. (2006). ECG mea-

surement on a chair without conductive contact. IEEE

transactions on bio-medical engineering, 53(5):956–

Ottenbacher, J., Kirst, M., Jatobá, L., Huﬂejt, M., Gross-

mann, U., and Stork, W. (2008). Reliable motion ar-

tifact detection for ECG monitoring systems with dry

electrodes. In Engineering in Medicine and Biology

Society, Annual International Conference of the IEEE,

pages 1695 –1698.

Park, C., Chou, P. H., Bai, Y., Matthews, R., and Hibbs, A.

(2006). An ultra-wearable, wireless, low power ECG

monitoring system. 2006 IEEE Biomedical Circuits

and Systems Conference, pages 241–244.

Schumm, J., Axmann, S., Arnrich, B., and Tröster, G.

(2009). Automatic Signal Appraisal for Unobtrusive

ECG Measurements. In Proceedings of the Biosignal

Interpretation Conference.

Zimmerman, T. G., Smith, J. R., Paradiso, J. A., Allport, D.,

and Gershenfeld, N. (1995). Applying electric ﬁeld

sensing to human-computer interfaces. In Proceed-

ings of the SIGCHI conference on Human factors in

computing systems - CHI ’95, pages 280–287, New

York, New York, USA. ACM Press.

BIOSIGNALS 2011 - International Conference on Bio-inspired Systems and Signal Processing

458