Professional Documents
Culture Documents
3, MAY 2015
Abstract—A mobile device is presented for monitoring both res- a noncontact way allowing unobtrusive monitoring. The device
piration and pulse. The device is developed as a bendable/flexible includes a microcontroller for data processing and a Bluetooth
inlay that can be placed in a shirt pocket or the inside pocket of module for data transmission. The entire device is realized on a
a jacket. To achieve optimum monitoring performance, the device
combines two sensor principles, which work in a safe noncontact small printed circuit board (PCB) which can easily be placed in
way through several layers of cotton or other textiles. One sensor, a shirt pocket or the inside pocket of a jacket. Since the carrier
based on magnetic induction, is intended for respiratory moni- material of the circuit board is flexible, the device adapts its
toring, and the other is a reflective photoplethysmography sensor form to the thoracic surface. Due to its pocket-sized flexible
intended for pulse detection. Because each sensor signal has some construction, it is called “FlexPock.”
dependence on both physiological parameters, fusing the sensor
signals allows enhanced signal coverage. The first sensor method incorporated in the FlexPock de-
vice is the magnetic induction (MI) technique; this is based on
Index Terms—Noncontact monitoring, pulse, respiration, sensor the electromagnetic coupling between a single sensing coil and
fusion, wearable sensors.
thoracic tissue. This method was chosen due to its excellent
I. INTRODUCTION ability to monitor respiratory activity [1]. The use of MI mea-
surements for physiological activity monitoring was introduced
OME or telemonitoring systems need frequent records of
H vital signs on a regular basis to assess the health status of
a patient. To maintain the patients’ quality of life, monitoring of
in 1967 by Vas et al. and called “displacement cardiograph”
[2] (Wilson et al. later claimed this device to work solely via
capacitive coupling [3]). Over the last decades, this method has
vital signs should take place as unobtrusively as possible. For been sporadically investigated by various groups in stationary
this purpose, on-body sensors can be of considerable benefit. setups [4]–[7]. In 2014, a mobile textile-integrated MI device
An ideal on-body sensor for home application should be mo- was presented by our group [8].
bile and easily wearable, so as not to restrict the patient’s mo- The use of the second sensor method is intended for pulse
bility. Also, it should be easy to use without the need for skilled measurement. For this purpose, an optical sensor was designed
personal and/or complex electrode application on multiple mea- based on photoplethysmography, a technique that was intro-
surement locations. Because the device should be suitable for duced in 1935 [9]. Photoplethysmographic sensors emit light of
long-term monitoring, direct skin contact should be avoided to a specific wavelength into the tissue region under investigation,
prevent skin irritation. Finally, for better acceptance by the pa- measure the amount of light that passes through the tissue, and
tients, the sensor should be imperceptible, i.e., light weight, flat, arrive at a measurement unit. Since the light intensity at the mea-
and adaptive to body motion. surement unit depends on blood content in the tissue, this sensor
The novel device presented here for monitoring respiration technique is well suited for cardiac pulse detection. The sensor
and pulse meets all the aforementioned requirements of an ideal developed for the present device is a reflective photoplethysmo-
on-body sensor. It combines two sensors, both of which work in graph [10], i.e., the light source and light measurement unit are
on the same side of the tissue, and the fraction of light reflected
Manuscript received October 31, 2014; revised February 1, 2015; accepted
March 20, 2015. Date of publication March 27, 2015; date of current version
(after superficial penetration) by the thorax is recorded. A sim-
May 7, 2015. ilar method was applied by our group for the development of a
D. Teichmann is with the Philips Chair for Medical Information Technology, wearable in-ear sensor for oxygen saturation monitoring [11].
RWTH Aachen University, 52074 Aachen, Germany (e-mail: teichmann@
hia.rwth-aachen.de).
Combining both sensor techniques enables us to monitor vari-
D. De Matteis is with RWTH Aachen University, Aachen 52074, Germany ous physiological parameters at the same measurement location.
(e-mail: dennis.de.matteis@rwth-aachen.de). This reduces both application effort and the size of the device,
T. Bartelt was with the Philips Chair for Medical Information Technology,
RWTH Aachen University, 52074 Aachen, Germany, when this work was carried
and allows us to investigate the dependence between different
out. Currently he is with Fritz Stephan GmbH, 56412 Gackenbach, Germany physiological measures without time-shifts or damping effects
(e-mail: thorsten.bartelt@rwth-aachen.de). due to mechanical propagation.
M. Walter and S. Leonhardt are with the Philips Chair for Medical Informa-
tion Technology, RWTH Aachen University, 52074 Aachen, Germany (e-mail:
Section II presents an overview of the FlexPock device. The
walter@hia.rwth-aachen.de; leonhardt@hia.rwth-aachen.de). physical principle and technical realization of the MI and re-
Color versions of one or more of the figures in this paper are available online flective photoplethysmographic sensor are described in Sections
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JBHI.2015.2417760
II-B and II-C, respectively. To enable mobile operating of the
2168-2194 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
TEICHMANN et al.: BENDABLE AND WEARABLE CARDIORESPIRATORY MONITORING DEVICE FUSING TWO NONCONTACT SENSOR 785
B. rPPG Sensor
1) Physical Principle: Human skin is an inhomogeneous
Fig. 1. (a) Block diagram of the system and (b) photograph of the FlexPock
device. and anisotropic optical medium. Light of a specific wavelength
is absorbed, reflected, or transmitted by the skin.
The reflectivity of the skin, i.e., the fraction between reflected
device with a single lithium polymer (LiPo) battery, an elaborate and incident light intensity, is correlated with the amount of
power management was implemented (which is described blood within the subcutaneous tissue. Reflectivity also depends
in Section II-D). Laboratory experiments and finite element on oxygen saturation of the blood. This dependence is utilized
method (FEM) simulations were conducted to optimize and for the estimation of blood oxygenation. For this purpose, light
characterize the sensors (see Sections III-A and III-B). Finally, of at least two different wavelengths is typically used.
a first proof of concept was applied in four healthy volun- For detection of the pulse rate, it is sufficient to use light of
teers and the results of these measurements are presented in a single wavelength emitted by LEDs and observe the alternat-
Section III. Preliminary information on this device and parts of ing component of the reflected (or transmitted) light by means
this paper were already presented in [12]. of a photodiode (see Fig. 2). When the heart pumps blood to
the periphery during a cardiac cycle, the pressure pulse reaches
II. METHODS the subcutaneous tissue within a certain time lag and modulates
the photoplethysmographic signal. The height of the signal am-
A. System Overview plitude measured by the photodiode is proportional to the dif-
Fig. 1(a) presents a system overview of the FlexPock device ference between systolic and diastolic pressure. The constant
in the form of a block diagram. The device uses a MI sensor for component is almost entirely caused by the basic absorption of
respiratory measurement and a reflective photoplethysmography the observed tissue.
(rPPG) sensor for pulse measurement. As already mentioned in Besides pulse activity, other physiological processes may also
Section I, the MI sensor is based on electromagnetic coupling influence the rPPG signal. Especially, respiration can vary the
and should not be interpreted as a sensor for magnetic field flux subcutaneous reflectivity by affecting the amount of blood vol-
density. The sensor signals are collected by a microcontroller ume pumped by the heart, as well as by increasing the pressure
(MSP430F5437A, Texas Instruments) where data are processed on thoracic tissue.
and passed to a Bluetooth module (BlueMod+B20/BT2.1, Stoll- 2) Realization: Fig. 2 shows the block diagram of the pho-
mann) for wireless transmission to a display unit. The display toplethysmographic sensor developed for the FlexPock device.
unit can be a personal computer (running C++ Software with To ensure that enough light penetrates the textile layer in front
QT-Library) or any Android device (running a Java App). of the FlexPock device, the illumination of the subcutaneous
786 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 19, NO. 3, MAY 2015
Fig. 4. Power management of the FlexPock device. Fig. 5. Pulse measurements with different setups of the rPPG sensor. (a) Two
LEDs, LED-to-photodiode distance 20 mm. (b) Two LEDs, LED-to-photodiode
distance 15 mm. (c) One LED, LED-to-photodiode distance 15 mm.
TABLE I TABLE II
PARAMETERS OF THE TESTED COILS SNR OF THE COIL TEST MEASUREMENTS PRESENTED IN FIG. 6
Fig. 7. Simplified model of the thorax for FEM simulation. Coil is drawn in
red.
the results of coil no. 1 with those of coil no. 3 shows that the
coil’s radius also has a significant effect on the sensor signal:
Although coil no. 3 has a higher inductivity, coil no. 2 with the
greater coil area produces a better respiratory signal.
According to these findings, the coil for the final FlexPock
design was chosen to have an outer radius of 30 mm [which is
close to the maximum size that fits into a standard shirt pocket
Fig. 6. Respiratory measurements with different setups of the MI sensor. (a) (approx. 100 mm)] and five windings (providing enough space
n = 3, ri = 25 mm. (b) n = 5, ri = 25 mm. (c) n = 8, ri = 5 mm.
within the coil for the components of the rPPG sensor).
distance of 15 mm should not be exceeded, since even 5 mm B. FEM Simulation of the Induced Current Density and its
additional distance reduces the pulse amplitude by half. Dependence on Coil Deformation
As each LED uses 100 mA, it would be beneficial for energy To obtain further information on the physical performance
saving to reduce the number of LEDs. For this reason, pulse of MI for thoracic monitoring, simulations based on the FEM
measurement at the wrist was also conducted with only one were conducted. In this way, the impact of coil deformation as
LED. The results of this experiment are presented in Fig. 5(c) expected for a bendable measurement device could be inves-
and show that the use of fewer LEDs significantly reduces the tigated. FEM simulations were done using the ac/dc module
signal amplitude and that the use of multiple LEDs is therefore of the COMSOL multiphysics software package (Comsol Inc.,
recommended. Burlington, USA).
2) Coil Design: Three different coils were produced and A simplified thorax model shown in Fig. 7 was composed
tested for their suitability for respiratory monitoring. They were of simple three-dimensional geometries. To avoid mathematical
connected with a MI sensor (as described in Section II-C2) and convergence problems, the model was symmetrically arranged.
placed on the chest by means of a flexible belt. The parameters The thorax is represented by an ellipsoid embedded in a sphere
of the different coils are presented in Table I. Two coils with the representing the surrounding air. The coil comprises one turn of
same inner radius, but different windings, were tested as well copper (diameter 1 mm) and an outer diameter of 60 mm. It is
as a coil with a smaller radius but an inductivity with a range driven by an alternating current of Icoil = 1 mA and a frequency
similar to that in the other two coils. The offset compensated of 30 MHz. Between the coil and thoracic wall, there is an air
results are shown in Fig. 6. The SNR of each measurement is gap of 1 mm. Table III summarizes the geometric dimensions
presented in Table II. The SNR was calculated by the ratio of the and material properties of the different organs. The electrical
signal (≤ 5 Hz) and high-frequency noise (≥ 5 Hz) in decibels. properties of the organs were taken from [13].
The higher inductivity of coil no. 2 (n = 5, L = 2.85 μH) In 1968, Tarjan and McFee [4] claimed that MI recordings of
yields to much better signal than achieved with coil no. 1 of the heart show the best signal quality during maximum inspira-
the same size but with lower inductivity (n = 3, L = 1.44 μH). tion. The authors assumed that due to the distal displacement of
The SNR of coil no. 2 is about 44.37 dB higher. Comparing the diaphragm and the lower conductivity of the inflated lung,
TEICHMANN et al.: BENDABLE AND WEARABLE CARDIORESPIRATORY MONITORING DEVICE FUSING TWO NONCONTACT SENSOR 789
Fig. 10. Magnetic flux density (B-field) in arrow presentation on the transver-
sal (xy) plane during coil elongation in x-direction by factor (a) 2 and (b) 2.5.
For physical dimensions, see Fig. 7 and Table III. Overlapping regions between
heart and lungs are assigned to heart tissue.
Fig. 8. Absolute value of the induced current density on the frontal (xz) plane
positioned at the center of the heart during (a) expiration (smaller volume and
higher conductivity of the lungs) and (b) inspiration (higher volume and lower
conductivity of the lungs). For physical dimensions, see Fig. 7 and Table III.
Fig. 11. Absolute value of the induced current density on the frontal (xz)
plane positioned at the center of the heart during (a) convex and (b) concave
Fig. 9. Absolute value of the induced current density on the frontal (xz) plane coil deformation. For physical dimensions, see Fig. 7 and Table III.
positioned at the center of the heart with the coil translated by 0, 20, 40, and
60 mm. For physical dimensions, see Fig. 7 and Table III.
the heart turns more into the focus of the measurement coil. To
validate this hypothesis by means of the FEM facilities available
today, the distribution of the induced current density during lung
expiration and inspiration was simulated and are compared in
Fig. 8(a) and (b), respectively. The intersecting plane lies in the
xz plane (i.e., frontal plane) at the middle of the heart (on the
y-axis).
Both simulations show a distribution of the induced current Fig. 12. Healthy volunteer with the FlexPock device.
density which is concentric around the coil center and has its
maximum value at the heart’s surface. Due to the higher con-
ductivity of the expired lung, there are high current densities in still mainly focuses in the heart region. In fact, the calculated
the direct surrounding of the heart and the transition between reflected impedance of the coil (which is given for each dis-
the organ boundaries is not well pronounced. In contrast, dur- placement step in Table IV) shows a higher value when shifted
ing inspiration, a much more pronounced change in the current by 20 mm than with the coil at the central position. This may
density distribution is visible, which is caused by the higher indicate that the heart has more impact on the coil’s impedance
conductivity of the lungs. at this slightly translated measurement location. When the coil
Fig. 9 shows the effect of lateral displacement of the coil. The is further translated to the side (by 40 and 60 mm), the induced
displacement starts at a central position directly above the heart eddy currents inside the heart decrease in favor of less con-
and comprises three consecutive translations by 20 mm. As can ductive tissue regions; this causes the strong decrease of the
be seen, the current density with the coil shifted by 20 mm reflected coil impedance, as shown in Table IV.
790 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 19, NO. 3, MAY 2015
Fig. 13. Measurement results on a healthy volunteer (subject 4 in Table V) with the FlexPock device placed on the left breast (inside a shirt pocket) during (a)
standing and (b) and (c) sitting posture. In (c), two additional layers of cotton were placed between skin and sensor (total of three layers). (a) Standing posture
(one layer of cotton between skin and sensor). (b) Sitting posture (one layer of cotton between skin and sensor). (c) Sitting posture (three layers of cotton between
skin and sensor).
Because the FlexPock device is flexible, the coil can become C. Monitoring of Respiration and Pulse
deformed during measurement. Therefore, the impact of coil To verify the ability of the device to monitor respiration and
deformation was investigated, i.e., elongation of the coil as well
pulse, four healthy male volunteers wore the FlexPock inside
as a convex (due to a drape of the shirt) or concave (generated their left-shirt pocket (see Fig. 12).
by motion of the thoracic wall due to respiration) coil curvature.
The volunteers were asked to perform 60 s of normal breath-
The terms convex and concave are here defined as referring to
ing as well as a 10-s apnea phase in both standing and sitting
the thorax. position. To investigate the device’s performance when more
Fig. 10 shows the magnetic flux density (B-field) in an arrow
than one thin textile layer is placed between the device and the
presentation on the transversal plane for a coil elongated in the
thorax, the experiment (in sitting position) was also conducted
x -direction by a factor 2 [see Fig. 10(a)] and by a factor 2.5 [see with volunteers wearing two cotton T-shirts under the shirt. The
Fig. 10(b)]. The length and thickness of an arrow represent the
derived signals were compared to simultaneously recorded res-
field strength at the position of the arrow’s shaft. Apparently,
piratory (Flowmeter, Model 4040, TSI Inc.) and cardiac (Elec-
the B-field smears over, the more the coil is elongated. This trocardiogram, IntelliVue MP70, Philips GmbH) references.
implies a lower penetration depth into tissue and less focus in
To evaluate the quality of the derived signals, three perfor-
the direction of elongation. mance metrics were calculated for each sensor.
Fig. 11 shows the induced current densities in the frontal plane (1) Respiration-to-pulse ratio (RPR): The ratio between the
for a convex [see Fig. 11(a)] and a concave [see Fig. 11(b)] coil
mean peak-to-peak amplitude value of the respiratory and
curvature. In the case of a convex curvature, two centers of cardiac cycles.
high current density arise, whereas in the case of concave cur- (2) SNR of the cardiac (SNRpulse ) and respiratory (SNRresp )
vature, the current density distribution maintains the concentric
signal content: The ratio between the mean peak-to-peak
characteristic of the nondeformed coil.
TEICHMANN et al.: BENDABLE AND WEARABLE CARDIORESPIRATORY MONITORING DEVICE FUSING TWO NONCONTACT SENSOR 791
IV. DISCUSSION
The design of the MI sensor and the rPPG sensor conforms
to the results presented in Section III-A. The signal quality
of the MI sensor increases with both the coil’s radius and its
inductivity. Therefore, the coil’s outer radius was chosen to
be close to the maximum size that fits into a standard shirt
pocket, while the inner radius provides enough space inside the
coil for the components of the rPPG sensor. According to the
results in Section III-A1, the distance between the LED and the
photodiode was chosen as close as possible.
The results in Section III-C illustrate the ability of the device
to adequately monitor cardiorespiratory activity. Nevertheless,
a more extensive evaluation of the device with more measure-
ments derived from more volunteers is needed. In particular, the
effect of motion artifacts has to be explored. To determine to
what extent and how often the device can be bent without loss
of soldering quality, endurance testing has to be applied to the
FlexPock device.
At the current development stage, the operation period of
the battery is a limiting factor. The operation period of the de-
vice could be increased by reducing the sampling as well as
the Bluetooth transmission rate (a low-power Bluetooth trans-
In (c) two additional layers of cotton were placed between skin and sensor (total of mission stack is also recommended). Furthermore, the LEDs
three layers). of the rPPG sensor could be pulsed when the signal from the
photodiode is digitized. Since the LEDs need 330 mW in total,
this procedure would dramatically decrease the device’s power
amplitude value of the cardiac or respiratory cycles and consumption. In this way, the effect of ambient light could also
the noise floor (two times the root-mean-square value) be compensated for by performing measurements without LED
in decibels (dB). Noise was defined as all signal content light and subtracting it from the measured signal.
above 5 Hz. The peak-to-peak values of the cardiac cycles Both respiratory and cardiac activity were contained in the
were measured during apnea phase. signal of the MI sensor and the rPPG sensor when placing the
Fig. 13 shows the representative excerpts of measurements FlexPock device on the left chest. The much higher respiratory
(three respiratory cycles and a 5-s apnea phase) recorded from signal content overlays the lower cardiac one and, therefore,
one of the volunteers. Note that the respiratory reference pro- complicates pulse detection. Since the RPR was much lower for
vides the absolute value of the respiratory flow; therefore, the rPPG sensor, this sensor is an ideal complement to the MI
two amplitude waves correspond to one respiratory cycle. Ta- sensor which, on the other hand, provides an excellent respira-
ble V provides the calculated performance metrics of the tory signal.
three measurements for each volunteer, as well as their average Since the MI sensor is not restricted to optical coupling,
values. its SNRpulse value shows only a low decrease (ΔSNRpulse =
Apparently, in standing posture [see Fig. 13(a)] both sensors −8.6 dB) in comparison to the one of the rPPG sensor
provide excellent SNR values for respiration and pulse monitor- (ΔSNRpulse = −32.8 dB) when additional layers of cotton tex-
ing. The RPR of the rPPG signal is much lower than that of the tile are placed between the FlexPock device and skin. The in-
MI signal. This difference in the sensors’ RPR values is even crease of RPR in the rPPG sensor case due to additional textile
more pronounced in the sitting position, which is also reflected layers can be explained by the decrease of optical coupling and,
by the increase in values of the MI sensor’s SNRresp and the hence, a higher relative influence of respiratory motion.
rPPG sensor’s SNRpulse (see Table V). Measurement locations other than the left pectoralis muscle
When three layers of cotton textile are placed between the (i.e., shirt pocket) could also be advantageous. For instance, a
FlexPock device and the thoracic skin, the respiratory and car- measurement location on the back of the thorax might yield to a
diac signals are still detectable [see Fig. 13(c)] but show a sig- higher RPR of the MI sensor due to the increased distance from
nificant loss of signal quality. the heart. Furthermore, the back of the thorax generally shows
792 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 19, NO. 3, MAY 2015
less motion [14]; this could decrease the signal content obtained [6] R. Guardo, S. Trudelle, A. Adler, C. Boulay, and P. Savard, “Contactless
by the rPPG sensor due to respiratory motion. recording of cardiac related thoracic conductivity changes,” in Proc. IEEE
Ann. Int. Conf. Eng. Med. Biol. Soc., Quebec, Canada, Sep. 20–24, 1995,
The fact that both sensors measure both respiration and pulse vol. 2, pp. 1581–1582.
could be used to enhance the coverage of respiratory and pulse [7] J. Vedru and R. Gordon, “Model of an inductive sensor of cardiac activity
rate estimation, i.e., the time during which a parameter ex- attached to patient,” J. Phys., Conf. Ser., vol. 224, 2010, doi: 10.1088/1742-
6596/224/1/012010.
traction is possible. The fusion of two sensors at the same [8] D. Teichmann, A. Kuhn, S. Leonhardt, and M. Walter, “The MAIN shirt:
measurement location offers additional advantages: It allows A textile-integrated magnetic induction sensor array,” Sensors, vol. 14,
to investigate the dependence between different physiological no. 1, pp. 1039–1056, 2014.
[9] K. Matthes, “Untersuchungen über die sauerstoffsättigung des men-
measures without time lag or mechanical damping caused by schlichen arterienblutes,” Naunyn-Schmiedebergs Archiv für Experi-
different measurement locations. Such a measure could be the mentelle Pathologie und Pharmakologie, vol. 179, no. 6, pp. 698–711,
time-interval between the ejection of the heart (measured by the 1935.
[10] K. Li and S. Warren, “A wireless reflectance pulse oximeter with dig-
MI sensor via cardiac wall motion) and the corresponding ar- ital baseline control for unfiltered photoplethysmograms,” IEEE Trans.
rival of the blood volume in the subcutaneous tissue (measured Biomed. Circuits Syst., vol. 6, no. 3, pp. 269–278, Jun. 2012.
by the rPPG sensor). The spatial fusion of both sensors also al- [11] B. Venema, N. Blanik, V. Blazek, H. Gehring, A. Opp, and S. Leonhardt,
“Advances in reflective oxygen saturation monitoring with a novel in-ear
lows the possibility of motion artifact cancellation, since motion sensor system: Results of a human hypoxia study,” IEEE Trans. Biomed.
artifacts will presumably couple in both sensors simultaneously Eng., vol. 59, no. 7, pp. 2003–2010, Jul. 2012.
and to the same extent. However, further verification of these [12] D. Teichmann, D. D. Matteis, M. Walter, and S. Leonhardt, “A bendable
and wearable cardiorespiratory monitoring device fusing two noncontact
ideas has to be provided in future investigation. sensor principles,” in Proc. 11th Int. Conf. Wearable Implantable Body
Sensor Netw., Zurich, Switzerland, Jun. 16–20, 2014, pp. 58–63.
[13] D. Andreuccetti, R. Fossi, and C. Petrucci. (1997). An Internet Resource
V. CONCLUSION for the Calculation of the Dielectric Properties of Body Tissues in the
Frequency Range 10 Hz–100 GHz (Based on data published by C. Gabriel
The device presented here shows excellent ability to moni- et al. in 1996). [Online]. Available: http://niremf.ifac.cnr.it/tissprop/
tor cardiorespiratory activity. Despite several layers of cotton [14] A. D. Groote, M. Wantier, G. Cheron, M. Estenne, and M. Paiva, “Chest
wall motion during tidal breathing,” J. Appl. Physiol., vol. 83, no. 5,
textile between the sensor and skin, it is still possible to ob- pp. 1531–1537, 1997.
tain signals suitable for the extraction of respiratory and pulse
rate (MI sensor: SNRresp = 98.5 dB, SNRpulse = 42.2 dB; rPPG
sensor: SNRresp = 62.3 dB, SNRpulse = 27.7 dB). Combining
two noncontact sensor principles and placing them at the same Daniel Teichmann (S’12–M’15) was born in Essen,
measurement location allows enhancement of both physiologi- Germany, in 1982. He received the Dipl.Ing. degree in
electrical engineering from RWTH Aachen Univer-
cal information and signal quality. It has been shown that it is sity, Aachen, Germany, where he is currently working
likely to happen that the amount of cardiac or respiratory related toward the Dr. Ing. degree with the Chair of Medical
signal content of the MI and rPPG sensor changes in dependence Information Technology.
He is currently a Research Assistant at RWTH
on body posture. Therefore, by fusing both sensor signals, the Aachen University. His research interests include
coverage rate of the parameter extraction could be enhanced. noncontact monitoring techniques and signal pro-
Furthermore, physiological measures derived by combinations cessing.
of both signals could be monitored. The spatial sensor fusion
enables the use of adaptive motion artifact cancellation tech-
niques because both signals will be affected by the same motion
artifact. Since the device is mobile, wearable, easy to apply, easy Dennis De Matteis was born in Hagen, Germany,
in 1985. He is currently working toward the M.Sc.
to operate, noncontact, unobtrusive, motion adaptive, and mul- degree in computer engineering from RWTH Aachen
timodal, it seems to be well suited for on-body sensor networks University, Aachen, Germany.
in telemonitoring applications.
REFERENCES
[1] D. Teichmann, J. Foussier, J. Jia, S. Leonhardt, and M. Walter, “Noncon-
tact monitoring of cardiorespiratory activity by electromagnetic coupling,”
IEEE Trans. Biomed. Eng., vol. 60, no. 8, pp. 2142–2152, Aug. 2013.
[2] R. Vas, “Electronic device for physiological kinetic measurements and
detection of extraneous bodies,” IEEE Trans. Biomed. Eng., vol. BME-14,
no. 1, pp. 2–6, Jan. 1967.
[3] D. L. Wilson and D. B. Geselowitz, “Physical principles of the displace- Thorsten Bartelt was born in Viersen, Germany,
ment cardiograph including a new device sensitive to variations in torso in 1982. He received the Dipl.Ing. degree in elec-
resistivity,” IEEE Trans. Biomed. Eng., vol. BME-28, no. 10, pp. 702–710, trical engineering from RWTH Aachen University,
Oct. 1981. Aachen, Germany.
[4] P. P. Tarjan and R. McFee, “Electrodeless measurements of the effective He is currently working with “Fritz Stephan
resistivity of the human torso and head by magnetic induction,” IEEE GmbH” at Gackenbach, Germany.
Trans. Biomed. Eng., vol. BME-15, no. 4, pp. 266–278, Oct. 1968.
[5] M. G. Pepper, D. J. E. Taylor, and M. C. Kwok, “Noninvasive detection of
ventricular wall motion by electromagnetic coupling: Part 2: Experimen-
tal,” Med. Biol. Eng. Comput., vol. 29, no. 2, pp. 141–148, 1991.
TEICHMANN et al.: BENDABLE AND WEARABLE CARDIORESPIRATORY MONITORING DEVICE FUSING TWO NONCONTACT SENSOR 793
Marian Walter (M’97–SM’13) was born in Steffen Leonhardt (M’95–SM’06) was born in
Saarbrücken, Germany, in 1966. He studied electrical Frankfurt, Germany, in 1961. He received the M.S.
engineering, with a specialization in control engineer- degree in computer engineering from the State Uni-
ing, and received the Dipl.Ing. and Dr. Ing. degrees versity of New York, Buffalo, NY, USA in 1987, the
from Technical University of Darmstadt, Darmstadt, Dipl.Ing. degree in electrical engineering, in 1989
Germany, in 1995 and 2002, respectively. and the Dr. Ing. degree in control engineering from
He was with medical engineering industry for the Technical University of Darmstadt, Germany, in
three years and was appointed as a Senior Scientist 1995, and the M.D. degree in medicine from J. W.
and the Deputy Head at the Philips Chair of Medical Goethe University, Frankfurt, Germany, in 2001.
Information Technology at RWTH Aachen Univer- He has five years of R&D management experience
sity, Aachen, Germany, in 2004. His research inter- in medical engineering industry and was appointed as
ests include noncontact monitoring techniques, signal processing, and feedback a Full Professor and the Head of the Philips endowed Chair of Medical Informa-
control in medicine. tion Technology at RWTH Aachen University, Germany, in 2003. Among others,
Dr. Leonhardt serves as an associate Editor of the IEEE Journal of Biomedi-
cal and Health Informatics and IEEE Transactions on Biomedical Circuits and
Systems. In 2014, he became a fellow of the NRW Academy of Sciences, Hu-
manities and the Arts in Düsseldorf. In 2015, he was appointed a distinguished
lecturer by the EMBS.