52

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 2, NO. 1, MARCH 2012

VLSI Friendly ECG QRS Complex Detector

for Body Sensor Networks
Chris F. Zhang and Tae-Wuk Bae

AbstractThis paper aims to present a very-large-scale integration (VLSI) friendly electrocardiogram (ECG) QRS detector for
body sensor networks. Baseline wandering and background noise
are removed from original ECG signal by a mathematical morphological method. Then the multipixel modulus accumulation is
employed to act as a low-pass filter to enhance the QRS complex
and improve the signal-to-noise ratio. The performance of the algorithm is evaluated with standard MIT-BIH arrhythmia database
and wearable exercise ECG Data. Corresponding power and area
efficient VLSI architecture is designed and implemented on a commercial nano-FPGA. High detection rate and high speed demonstrate the effectiveness of the proposed detector.
Index TermsBody sensor networks (BSNs), electrocardiogram
(ECG) sensor, field-programmable gate array (FPGA), mathematical morphology, QRS detection, very-large-scale integration
(VLSI) architecture.

I. INTRODUCTION

HE last decade has witnessed a rapid surge of interest in

new sensing and monitoring devices for healthcare and
the use of wearable/wireless devices and sensor networks for
clinical applications. Especially, a new wave of technology,
body sensor networks (BSNs), has played an increasingly
important role in providing continuous diagnosis support and
medical treatment.
As one of the important physiological sensor nodes in BSN,
wearable electrocardiogram (ECG) sensor is dedicated to measuring the rate and regularity of heartbeats as well as the size
and position of the chambers, the presence of any damage to
the heart and the effects of drugs or devices used to regulate
the heart [1]. In ECG signal processing, all the extensive analysis need the information of QRS positions as a basic [2][4].
QRS detectors have been regarded as a mature topic until the
BSN is introduced, where, unfortunately, the ECG sensor requires real-time, miniature form factors and long lifetimes that
push the limits of ultra low power circuit and system design.
Among the noises plaguing the ECG are the power-line interference: 50/60 Hz pickup and harmonics from the power cord;
electrode contact noise: baseline drift due to variable contact
between the electrode and the skin; motion artifacts: shifts in
Manuscript received October 07, 2011; revised January 06, 2012; accepted
January 30, 2012. Date of publication March 07, 2012; date of current version
April 11, 2012. This paper was recommended by Guest Editor H.-J. Yoo.
C. F. Zhang is with the Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, PA 15213 USA (e-mail:
chrisfzhang@gmail.com).
T. Bae is with the Stanford Center for Image Systems Engineering, Stanford
University, Stanford, CA 94305 USA.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JETCAS.2012.2187706

the baseline caused by changes in the electrode-skin impedance;

muscle contraction: electromyogram-type signals (EMG); respiration causing drift in the baseline; and electromagnetic interference and noise coupled from other electronic devices. For
meaningful and accurate detection, steps have to be taken to
filter out or discard all these noise sources. Hence, a reliable
on-the-fly QRS detection method with low hardware cost, high
sensitivity, and good noise susceptibility is of urgent need.
The front end of an ECG sensor should be able to deal
with extremely weak signals ranging from 0.5 to 5.0 mV,
mV and a
usually mixed with a dc component of up to
common- mode component of up to 1.5 V resulting from the
electrode- skin contact and the potential between the electrodes
and ground, respectively [1]. Depending on the specific application, the useful bandwidth of an ECG signal can range from 0.5 to
50 Hz for general healthcare purposes. A standard clinical ECG
application utilizes a bandwidth of 0.05 to 100 Hz. While for a
monitoring application in intensive care units, it could reach up
to 1 kHz for late-potential measurements (pacemaker detection).
Much efforts have been given to the frequency-based ECG filtering algorithms for QRS detection, i.e., a band-pass filter with
a center frequency in the range of 1017 Hz. After passing the
filter, the signal may be squared or averaged over a number of
samples to obtain the place of QRS waves. But these techniques
suffer from the fact that frequency bands of the noise/other components such as P/T waves overlap with that of QRS complex. In
addition, in order to accurately detect the QRS, prior knowledge
of frequency spectrums of different components in the recorded
ECG signal is required through an additional training process.
While the BSN devices are generally restricted by size, power
consumption, and computational load, it will be worthwhile exploring a simple and accurate QRS detection algorithm based
on morphology, instead of frequency, for the BSN ECG devices. Following this, in this paper, we utilize two most fundamental morphological operators (erosion and dilation) [5], [6]
in mathematical morphology technology to reduce the computational complexity for wearable ECG QRS detector. A power
and size efficient very-large-scale integration (VLSI) architecture is designed and implemented on field-programmable gate
array (FPGA) to verify the method.
This remainder of paper is organized as follows. In Section II,
a brief introduction of mathematical morphology filtering is
given, which serves as a basis for the proposed algorithm.
Section III presents the new algorithm and discusses the details
of the algorithm. What follows is the VLSI architecture design
and implementation in Section IV. The evaluation of the algorithm is done in Section V. Conclusion remarks are drawn in
Section VI.

2156-3357/$31.00 2012 IEEE

ZHANG AND BAE: VLSI FRIENDLY ECG QRS COMPLEX DETECTOR FOR BODY SENSOR NETWORKS

53

II. THEORY OF MATHEMATICAL MORPHOLOGY

Mathematical morphology is a set-theoretic method of image
analysis providing a quantitative description of geometrical
structures. And it provides an effective way to analyze signals
using nonlinear signal processing operators incorporating shape
information for extracting image components that are useful for
representation and description. A morphological operation is
actually the interaction of a set or function representing the object or shape of interest with another set or function of simpler
shape called structure element. The geometry information of
the signal is extracted by using the structure element to operate
on the signal. The shape of the structure element determines the
shape information of the signal that is extracted under such an
operation. Such operators serve two purposes, i.e., extracting
the useful signal and removing the artifacts.
There are two most basic morphological set transformation
operators: dilation and erosion, which all other mathematical
morphology operations are based on. The operators for 1-D
and structure element
are listed below for easy
signal
reference, i.e.,
(1)
(2)
where indicates the th element in a length structure element,
is a predefined structure element.
and
Opening and closing are two extended morphological operators based on dilation and erosion. In mathematical morphology,
opening is the dilation of the erosion of a set by a structuring element; the closing of a set by a structuring element is the erosion
of the dilation of that set. Opening and closing operations could
also work as morphology filters with clipping effects, i.e., cutting down peaks and filling up valleys
(3)
(4)
Another two common operators, top-hat and bottom-hat, encompass the opening and closing operators with two different
order, defined as
(5)
(6)
This morphology based method can effectively detect the
components of interest from contaminated signal without any
prior knowledge about frequency spectrum which embarrasses
the utilization of frequency based filters.
III. PROPOSED QRS DETECTION ALGORITHM
Detection of QRS complexes equals to the distinguishing of
a group of consecutive positive and negative peaks. As mentioned above, originating from 2-D image processing, mathematical morphological technology extracts the effective information based on shapes in the image, not pixel intensities like
conventional methods. A fundamental advantage of mathematical morphology applied to signal processing is that it is intuitive
since it works directly on the spatial domain: the structuring elements considered as the basic bricks play the same role as

Fig. 1. Block diagram of proposed algorithm.

frequencies do in the analysis of the frequently used frequency

filters. Another advantage is that it leads to better reproducible
results because of the strong mathematical foundation. The simplicity in terms of computation and hardware implementation
is also one of its main advantages. The last point that is worth
noting lies in its capability in allowing easy customization of the
program for the choice of every parameter according to physical
measurements if necessary.
The mathematical morphology was recently introduced into
1-D ECG processing [7][12], demonstrating the effectiveness
of removal of impulsive noise and background normalization.
However, existing [7][12] employed 8, 8, 8, 6, 4, and 2 high
order opening/closing operators or their computational combinations to realize the QRS detectors, respectively. In order to
further alleviate the implementation load and reduce the power
consumption, a VLSI friendly QRS detector only based on the
most fundamental dilation and erosion operator, as mentioned
in (1) and (2), is presented here, where only one dilation and
one erosion are employed, approximately equivalent to the computation complexity of one close or one opening operation according to (3) and (4). As shown below, this algorithm will offer
the lowest cost of hardware implementation in both power and
size when compared to [7][12].
The diagram of the proposed algorithm is shown in Fig. 1.
The proposed morphology operator filtering plays the most critical role in the proposed algorithm which removes the noise
and baseline drift and suppresses the P/T waves in ECG signal.
Then the multipixel modulus accumulation is used to enhance
the QRS complex. Finally, the threshold is applied to decide the
heart rate. The detailed discussions on each section in Fig. 1 are
presented in the following subsections. The standard MIT/BIH
ECG database [13] and our own wearable ECG data are used to
demonstrate the superiority of proposed QRS detector.
A. Morphological Filter
Now intuitively, dilation expands an image object and erosion shrinks it. The opening smoothes a contour in an image,

54

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 2, NO. 1, MARCH 2012

breaking narrow isthmuses and eliminating thin protrusions.

Closing tends to narrow smooth sections of contours, fusing
narrow breaks and long thin gulfs, eliminating small holes, and
filling gaps in contours. In most applications, opening is used
to suppress peaks while closing is used to suppress pits. Here,
in order to detect QRS complex accurately and quickly, a peak
extractor is defined only based on basic dilation and erosion
morphological operators, instead of a series of advanced openings and closings as in existing literature [7][12]. It can be
written as

(7)
For the ECG signal that processed by this morphology filter,
the QRS complex will be enhanced and others waves that
change smoothly will be mapped in horizontal axis after the
cancellation of dc components.
Reference [7] detailed how the structure element selection
is important for ECG QRS recognition in mathematical morphology, where qualitative morphology filtering investigations
were carried out for different structure elements applied to ECG
signal. Briefly, processing results, i.e., the signal information of
interest, can be mainly determined by three critical parameters
of the structure element in the morphological filter. The shape
(or slope) of the structure element is the most significant factor,
herein the smaller slope performs better in terms of removing
noise, but with a larger reduction in the amplitude of resulted
signal. The less significant item is the length of the structure element (i.e., the assumed duration of the QRS complex), which
varies with patient and placement of the electrodes, and longer
structure element turns out to lead to slightly better performance
in terms of noise reduction, but with a small decrease in amplitude of QRS complex. The least significant item is the amplitude of the structure item, which the filter performance is the
least sensitive to. Hence, in our selections, more considerations
are given in terms of the tradeoff between computational complexity, effectiveness, and amplitude of resulted signal when determining the parameters of structure element. Considering the
simplicity and similar shape, we select the triangle structure element by the optimization, defined as

(8)

where
denotes the smallest integer less than or equal
,
is the length of structure element,
to
and
is the amplitude. Here, we choose
, and
according to the above selection rules, hence the
.
B. Enhancing ECG by Modulus and Combination
The absolute value of the above output is then combined by
multiple-frame accumulation, which is much alike energy trans-

Fig. 2. Circuit diagram of the dilation/erosion module for a five-point

operation.

formation [14], [15]. The energy accumulation process is expressed as

(9)

The value of should correspond to the possible maximum

duration of normal QRS complex. This step further enhances
the filtered ECG signal to make QRS peaks easy to identify, as
shown in Figs. 315(d) in Section IV.
C. Threshold and Decisions
An adaptive threshold is used as the decision function in connection with the proposed transformation for QRS detection.
Usually, the threshold levels are computed signal dependent
such that an adaption to changing signal characteristics is possible. For the signal produced by (9), it is proposed that the required adaptive threshold is a function of the maximum of the
. The guideline in selecting the
transformed ECG waveform
threshold, , is given by
(10)
where Max is determined from the current signal segment which
is within the range of millivolts. The upper and lower bounds of
Max will be subject to the selection of structure elements. The
final results are shown in Figs. 315(e).
IV. VLSI ARCHITECTURE AND IMPLEMENTATION
The ultra-low power consumption is essential for ECG
sensor application in BSN. The straightforward format of (1),
(2), (7), and (9) facilitates the VLSI friendly implementation
of the proposed method. The similarity between dilation and
erosion operation further permits sharing the same architecture
of computation core. The design strategy of the hardware
implementation is to reduce as much computation load as
possible. As shown in Fig. 2, the proposed VLSI architecture
consists of shift registers, Rom, adder, comparator, and a finite
state machine (FSM)-based controller. For the convenience of
demonstration, the circuit realization of pivotal modules of a

ZHANG AND BAE: VLSI FRIENDLY ECG QRS COMPLEX DETECTOR FOR BODY SENSOR NETWORKS

Fig. 3. QRS detection for normal ECG using tape 100 of MIT/BIH database.

five-point length structure element

is
shown in Fig. 2 as an example. Here, the Reg means the shift
registers, storing the intermediate values, and Max/Min represents the comparators for dilation and erosion. The components
, mean the differences
for the Adder, i.e.,
between two values in the structure element, which are stored
and
in the Rom. In particular,
. This hardware design
strategy aims to reduce the number of adders for dilation and
erosion computation, since it is equivalent to comparing the
resulting items directly from the dilation and erosion even if
one constant number is added/subtracted before maximal and
minimum comparisons. In this fashion, whereas, only four
adders are needed, instead of five, resulting in 20% saving for
single channel. Under current design, the computation core
needs five subclock periods to finish one computation cycle: the
first one is to read five data from registers for adder, followed
by the clock period needed for the addition, and the next three
are subsequently used to compare the five resulting values from
the adders until the final result is output.
The opening and closing operations used in other literatures
[7][12] can be easily fulfilled by cascading the dilation and erosion units together. Hence, the proposed method outperforms
obviously in terms of implementation ease and computational
load. The rest parts of the circuits can be easily implemented
with a few shift registers and adders. The overall implementation of this algorithm is verified using the Actel nano-FPGA
starter kit. The FPGA operating under 3 kHz and 1.8 V supply
voltage can successfully detect the QRS complex as expected.
The power consumption is much lower than the FPGA [16] or
DSP [17] implementations of other solutions which dissipate at
least few milliwatts. The ultra low power feature of the proposed
QRS detection algorithm helps to extend the battery life, which
makes it extremely suitable for the BAN devices.
V. EXPERIMENTAL RESULTS
The MIT/BIH arrhythmia database is used for the evaluation
of the proposed QRS detection algorithm, as shown in Table I.
Figs. 38 give detection examples performed over different typical tapes from MIT/BIH database, where green and red markers

55

Fig. 4. QRS detection for large T wave using tape 114 of MIT/BIH database.

Fig. 5. QRS detection for large S wave using tape 117 of MIT/BIH database.

Fig. 6. QRS detection for pathological ECG using tape 200 of MIT/BIH
database.

are used to specify the true and detected QRS complexes, respectively. Fig. 4 gives a special case where the amplitude of
high T wave exceeds that of R wave. It clearly demonstrates
that the T wave will be eliminated after the mathematical morphology processing. It is intuitively challenging when the T
waves or noise spikes are similar in duration as the QRS complexes. How accurate this algorithm is will be a topic of interest

56

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 2, NO. 1, MARCH 2012

Fig. 7. QRS detection for irregular ECG using tape 207 of MIT/BIH database.

Fig. 10. Detection of QRS for our exercise ECG under running.

Fig. 8. QRS detection for using the tape 233 of MIT/BIH database.

Fig. 11. Detection of QRS for our exercise ECG under breathing deeply.

Fig. 9. Detection of QRS for our own exercise ECG under walking.

Fig. 12. Detection of QRS for our ECG under chest expanding exercise.

when the QRS duration is not sufficiently different from width

of the other waves or any noise spikes. Since it would be helpful
to show results of these corner cases for universal morphology
technology, active data collection and searching have been undergoing to obtain such ECG signals for evaluating the proposed

algorithm although it needs trial and error. To fully evaluate the

performance of detection algorithm, several indexes are introduced including false negative (FN) which means failing to detect a true beat, and false positive (FP) which represents a false
beat detection. By using FN and FP, the sensitivity (Se), positive

ZHANG AND BAE: VLSI FRIENDLY ECG QRS COMPLEX DETECTOR FOR BODY SENSOR NETWORKS

57

TABLE I
PERFORMANCE FOR PROPOSED ALGORITHM

Fig. 13. Detection of QRS for our ECG under standing up and crouching down.

Fig. 14. Detection of QRS for our ECG under going upstairs and downstairs.

Fig. 15. Detection of QRS for our ECG under jumping.

prediction
and detection error rate (DER) can be calculated using the following equations, respectively
%

(11)

(12)

(13)

58

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 2, NO. 1, MARCH 2012

TABLE II
COMPARISON WITH PUBLISHED ALGORITHMS

where true positive (TP) is the total number of QRS correctly

detected by the algorithm.
Based on the results of Table I, an average QRS detection rate
of 99.57%, a sensitivity of 99.76% and a positive prediction of
99.82% are obtained against the recordings of MIT/BIH database. Table II compares the existing algorithms with the proposed one by using tape 105, which is one of the most difficult
one for analysis due to large induced noises and is widely used
by researchers to test the QRS detection algorithm. It can be
seen, from the comparisons with the other mathematical morphology methods in [7], [11], and [12], that a comparable detection performance can be obtained while the proposed method
offers a considerable saving in computation load by only utilizing a small subset of operators, as described in Section III.
This fully demonstrates the proposed algorithm features the superior trade-off between the ease of VLSI implementation and
accuracy.
Figs. 915 show the detection results for wearable ECG data
collected by Welch Allyn PCE-210 PC-Based Exercise Stress
ECG System sampled at 600 Hz. They are extracted under
walking, running and going upstairs, respectively. It is obvious
that the proposed algorithm correctly detects the QRS of both
resting and exercise ECG, even under the presence of severe
noise, baseline drift and large P/T waves.
VI. CONCLUSION
This paper has presented a VLSI friendly morphology based
algorithm for both resting and wearable exercise ECG QRS detection in BSNs. Unlike the frequency based methods, the proposed method is free of the requirements of prior frequency
knowledge of signal component of interest. Also the method can
also work in case of the bandwidth overlaps between QRS complex and other components. The algorithm is implemented in
FPGA and evaluated with MIT/BIH standard ECG database to
achieve a QRS detection rate of 99.57%, a sensitivity of 99.76%
and a positive prediction of 99.82%. And it is also effectively
verified with our exercise ECG data with lower SNR. It is obvious that the proposed detector compares very favorably with
published results of other QRS detection algorithms.

REFERENCES
[1] G. Yang, Body Sensor Networks. New York: Springer, 2006.
[2] B. U. Kohler, C. Hennig, and R. Orglmeister, The principles of software QRS detection, IEEE Eng. Med. Biol. Mag., vol. 21, no. 1, pp.
4257, Jan. 2002.
[3] F. Zhang and Y. Lian, Novel QRS detection by CWT for ECG sensor,
in Proc. Int. Conf. IEEE Biomed. Circuits Syst., Montreal, ON, Canada,
Nov. 2007.
[4] H. Lee and K. M. Buckley, ECG data compression using cut and align
beats approach and 2D transforms, IEEE Trans. Biomed. Eng., vol. 46,
no. 5, pp. 556564, May 1999.
[5] A. Gasterators, I. Andreadis, and Ph. Tsalides, Fuzzy soft mathematical morphology, in IEEE Proc. Vis. Imag. Signal Process., Feb. 1998,
vol. 145, pp. 4149.
[6] F. Zhang, C. Li, and L. Shi, Detecting and tracking dim moving point
target in IR image sequences, Infrared Phys. Technol., vol. 46, pp.
323328, Apr. 2005.
[7] F. Zhang and Y. Lian, Effective ECG QRS detection based on multiscale mathematical morphology filtering, IEEE Trans. Biomed. Circuits Syst., vol. 3, no. 4, pp. 220228, Aug. 2009.
[8] J. Hu and S. Bao, An approach to QRS complex detection based on
multiscale mathematical morphology, in Proc. 3rd IEEE Int. Conf.
Biomed. Eng. Inform. (IEEE BMEI 2010), 2010, p. 725.
[9] C.-H. N. Chu and E. J. Delp, Impulsive noise suppression and background normalization of electrocardiogram signals using morphological operators, IEEE Trans. Biomed. Eng., vol. 36, no. 2, pp. 262273,
Feb. 1989.
[10] P. E. Trahanias, An approach to QRS complex detection using mathematical morphology, IEEE Trans. Biomed. Eng., vol. 40, no. 2, pp.
201205, Feb. 1993.
[11] Y. L. Chen and H. L. Duan, A QRS complex detection algorithm
based on mathematical morphology and envelope, in Proc. 27th Annu.
Int. Conf. IEEE EMBS, Shanghai, China, Sep. 2005, pp. 46544657.
[12] F. Zhang, J. Tan, and Y. Lian, An effective QRS detection algorithm for wearable ECG in body area network, in Proc. Int. Conf.
IEEE Biomed. Circuits Syst., Montreal, ON, Canada, Nov. 2007, pp.
195198.
[13] [Online]. Available: http://www.physionet.org/physiobank/database/
mitdb/
[14] N. V. Thakor, J. G. Webster, and W. J. Tompkins, Optimal QRS detector, Med. Bio. Eng. Comput., vol. 21, pp. 343350, May 1983.
[15] F. Gritzali, Towards a generalized scheme for QRS detection in ECG
waveforms, Signal Process., vol. 15, pp. 183192, Sep. 1988.
[16] T. Hoang, J. Son, Y. Kang, C. Kim, H. Chuang, and S. Kim, A low
complexity, low power, programmable QRS detector based on wavelet
transform for implantable pacemaker IC, in IEEE Int. SOC Conf., Sep.
2427, 2006, pp. 160163.
[17] A. Ruha and S. Sallinen, A real-time microprocessor QRS detector
system with a 1-ms timing accuracy for the measurement of ambulatory
HRV, IEEE Trans. Biomed. Eng., vol. 44, no. 2, pp. 159167, Feb.
1997.
[18] R. Poli, S. Cagnoni, and G. Valli, Genetic design of optimum linear
and nolinear QRS detectors, IEEE Trans. Biomed. Eng., vol. 42, no.
11, pp. 11371141, Nov. 1995.
[19] J. Pan and W. J. Tompkins, A real time QRS detection algorithm,
IEEE Trans. Biomed. Eng., vol. 32, no. 3, pp. 230236, Mar. 1985.
[20] S. Chen, H. Chen, and H. Chan, A real-time QRS detection method
based on moving-averaging incorporating with wavelet denosing,
Comput. Methods Programs Biomed., vol. 82, pp. 187195, Mar.
2006.
[21] V. X. Afonso, W. J. Tompkins, T. Q. Nguyen, and S. Luo, ECG beat
detection using filter banks, IEEE Trans. Biomed. Eng., vol. 46, pp.
192201, Feb. 1999.
[22] P. S. Hamilton and W. J. Tompkins, Quantitative investigation of QRS
detection rules using the MIT/BIH arrhythmia database, IEEE Trans.
Biomed. Eng., vol. 33, no. 3, pp. 11571165, Mar. 1986.
[23] Q. Xue, Y. H. Hu, and W. J. Tompkins, Neural-network-based adaptive filtering for QRS detection, IEEE Trans. Biomed. Eng., vol. 39,
no. 4, pp. 315329, Apr. 1992.
[24] J. Lee, K. Jeong, J. Yoon, and M. Lee, A simple real-time microprocessor QRS detection algorithm, in Proc. 18th Annu. Int. Conf. IEEE
Eng. Med. Biomed. Soc., 1996, pp. 13961398.

ZHANG AND BAE: VLSI FRIENDLY ECG QRS COMPLEX DETECTOR FOR BODY SENSOR NETWORKS

C. F. Zhang received the B.S. and Ph.D. degrees in

physics from Wuhan University, China, in 2002 and
2007, respectively.
Currently, he is working on the joint microwave
solid flow sensing project in National Energy
Technology Laboratory and the Department of
Electrical and Computer Engineering, Carnegie
Mellon University, Pittsburgh, PA. His research
area of interest includes semiconductor device,
microwave technology, and embedded systems, as
well as biomedical signal processing. He has been an
Editorial Committee Member for Journal of Circuits and Systems since 2010.
He has authored/coauthored around 80 peer-review papers and been awarded
two patents.
Dr. Zhang has been acting as a reviewer for many journals such as
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS,, IEEE TRANSACTION ON
BIOMEDICAL CIRCUITS AND SYSTEMS, IEEE TRANSACTION ON BIOMEDICAL
CIRCUITS AND SYSTEMS, IEEE TRANSACTIONS ON PLASMA SCIENCE, IEEE
TRANSACTIONS ON MECHATRONICS, IEEE POWER ELECTRONICS LETTERS, and
IEEE MICROWAVE AND WIRELESS COMPONENTS LETTERS. He has also been
involved in the international conferences as Conference Committee Member
and Review Committee Member such as ISCAS, ICMST, and BioCAS.

59

Tae-Wuk Bae received the B.S. and Ph.D. degrees

in electronics engineering from Kyungpook National University, Daegu, Korea, in 2004 and 2010,
respectively.
In 2006, he worked on an image quality enhancement as a research engineer in LG Display Co., Ltd.,
Korea. In 2011, he has been involved in a face recognition research as a postdoctoral researcher in Department of Electrical Engineering, KAIST (Korea Advanced Institute of Science and Technology), Korea.
Currently, he is researching a digital camera sensor
as postdoctoral researcher in SCIEN (Stanford Center of Image Engineering
Systems), Stanford University, Stanford, CA. He is a journal reviewer of Korea
Multimedia Society. His research interests are image, video signal processing,
color appearance model, and real-time embedded software.
Dr. Bae was a publication chair at the 11th Asian Conference on Computer
Vision.