Real-Time Compressed Sensing-Based Electrocardiogram
Compression on Energy-Constrained Wireless Body Sensors
Hossein Mamaghanian, Nadia Khaled, David Atienza and Pierre Vandergheynst
School of Engineering, Ecole Polytechnique Fédérale de Lausanne (EPFL)
Email: name.surname@epfl.ch
Abstract—Wireless body sensor networks (WBSN) hold the promise to
enable next-generation patient-centric tele-cardiology systems. A WBSN-
enabled electrocardiogram (ECG) monitor consists of wearable, minia-
turized and wireless sensors able to measure and wirelessly report cardiac
signals to a WBSN coordinator, which is responsible for reporting them
to the tele-health provider. However, state-of-the-art WBSN-enabled ECG
monitors still fall short of the required functionality, miniaturization and
energy efficiency. Among others, energy efficiency can be significantly
improved through embedded ECG compression, which reduces airtime
over energy-hungry wireless links. In this paper, we quantify the potential
of the emerging compressed sensing (CS) signal acquisition/compression
paradigm for low complexity energy-aware ECG compression on the
state-of-the-art ShimmerTM WBSN mote. Interestingly, our results show
that CS represents a competitive alternative to state-of-the-art digital
wavelet transform (DWT)-based ECG compression solutions in terms of
overall energy efficiency and ShimmerTM node lifetime extension.
I. I NTRODUCTION
The increasingly prevalent cardiac diseases are requiring escalating
levels of supervision and medical management, which are contribut-
ing to skyrocketing health care costs and, more importantly, are
unsustainable for traditional health care infrastructures. Wireless body
sensor networks (WBSN) promise to allow inexpensive, continuous
and remote health monitoring for next-generation of ambulatory per-
sonal tele-cardiology or mobile-cardiology systems. While the resting
electrocardiogram (ECG) monitoring is standard practice in hospitals,
its ambulatory counterpart is still facing many technical challenges.
Recently, the realization of wireless-enabled ultra-low-power ECG
monitors for ambulatory use has been receiving significant industrial
and academic interest [1], [2], [3]. Still, today’s state-of-the-art
ambulatory ECG monitors fall short in terms of either clinical
relevance and/or autonomy figure, primarily because they transmit
uncompressed ECG data over power-hungry wireless links. It is today
widely acknowledged that the achievement of truly WBSN-enabled
personal ECG monitoring systems requires more breakthroughs not
only in terms of ultra-low-power read-out electronics and radios, but
also in terms of embedded ECG compression and feature extraction
algorithms, and assorted ultra-low-power dedicated digital processors.
ECG compression relies on the sparse (and thus compressible)
nature of the ECG, as it can be approximated by a compact repre-
sentation in the wavelet domain [4]. Capitalizing on this sparsity, we
propose to apply the emerging compressed sensing (CS) approach [5],
[6], [7] for a low-complexity, real-time and energy-aware ECG signal
compression on WBSN motes. This is motivated by the observation
that this new signal acquisition and compression paradigm could be
well suited for low-power implementations as it dramatically reduces
the need for resource- (both processing and storage) intensive DSP
operations on the encoder side.
CS is a new sensing and processing paradigm, which challenges
the traditional analog-to-digital conversion based on the Nyquist-
Shannon sampling theorem. For sparse signals such as the ECG,
(above) Nyquist-rate sampling produces a large amount of redundant
digital samples, which are costly to wirelessly transmit and require
978-1-4244-9472-9/11/$26.00 ©2011 IEEE 1744
to be further compressed using non-linear digital techniques. CS
is a methodology that has been recently proposed to address this
problem. The main idea behind CS is relatively simple. Suppose
the considered signal has a (potentially large) bandwidth Ω but
has a sparse approximation, i.e., it can be represented by a linear
superposition of K elements of a dictionary. CS states that you only
need to collect roughly K samples (<< the number of samples
corresponding to Nyquist-rate sampling) using simple measurement
waveforms, thus sensing/sampling and compressing at the same time.
To the best of our knowledge, CS has never been applied to
ECG. It has, however, been recently considered for efficient EEG
acquisition and compression [8], [9], [10]. Accordingly, the present
work introduces 2 main contributions: (1) It is the first to thor-
oughly quantify the potential of CS for low-complexity energy-
aware ECG compression on resource-constrained WBSN platforms;
(2) To quantify this potential, it provides an exhaustive system-level
comparison between CS-based and state-of-the-art digital wavelet
transform (DWT)-based embedded ECG compression. This system-
level comparison is based on the embedded implementations of
the two considered ECG compression algorithms, which have been
optimized for real-time implementation on a representative state-
of-the-art WBSN mote, namely ShimmerTM [2]. In particular, our
comparative study of these 2 algorithms covers signal reconstruction
quality, embedded memory usage, CPU execution time and energy
consumption.
The rest of this paper is organized as follows. Section II introduces
the CS-based ECG compression and reconstruction scheme investi-
gated in this work, while Section III details the performance metrics
and the ECG database used for its performance evaluation. Then,
Section IV details our embedded implementation of CS-based ECG
compression. In Section V, its performance is compared to that of
state-of-the-art DWT-based compression. Finally, Section VI draws
the concluding remarks.
Notation: In all the following, normal letters designate scalar
quantities, boldface lower-case letters indicate column vectors, and
boldface capitals represent matrices. Moreover, mi and Mi,j are the
ith entry of vector m and the (i, j)th entry of matrix M, respectively.
Finally, (.)H and ||.||p denote the conjugate transpose, and the lp -
norm of a vector, respectively.
II. CS-BASED ECG C OMPRESSION
Figure 1 depicts the block diagram of CS-based ECG encoding.
Due to the limited on-chip memory and real-time constraints, the
digitized ECG signal is processed in non-overlapping windows of N
samples. Let x be the real-valued N -dimensional ECG signal vector
corresponding to such a window (i.e., x ∈ RN ). As aforementioned,
x has a sparse approximation, i.e., it can be represented by a
linear superposition of S elements of an orthonormal wavelet basis
Ψ = [ψ 1 |ψ 2 | · · · |ψ N ], as follows x = ΨαS , where αS represents
the N -dimensional coefficient vector with only S non-zero entries

Fig. 1. Block diagram of the CS-based ECG compression scheme imple-
mented on the ShimmerTM wireless mote
(S  N ). CS shows that it is sufficient to collect roughly S
samples using simple measurement waveforms. Accordingly, we
collect M samples using simple measurement vectors {φi }1≤i≤M as
yi = φ Hi x, i = 1, · · · , M . Consequently, the CS linearly compressed
data vector y ∈ RM is described by: y = Φx, where Φ denotes the
M ×N measurement or sensing matrix with the vectors φH
1 , · · · , φM
as rows. It is important to notice that the sensing matrix Φ does not
depend on the signal. To guarantee the robust and efficient recovery
of the S-sparse signal αS , the sensing matrix Φ must obey the key
restricted isometry property (RIP) [11]:
(1 − δS ) ||α||2 ≤ ||ΦΨα||2 ≤ (1 + δS ) ||α||2 ,
for all S-sparse vectors α. δS is the isometry constant of matrix Φ,
which must be not too close to one. A universal good choice for the
sensing matrix Φ are random matrices, such as random matrices with
independent identically distributed (i.i.d.) entries formed by sampling:
(1) a Gaussian distribution N√(0, 1/N ); (2) a symmetric Bernoulli
distribution (P (Φi,j = ±1/ N ) = 1/2). If the RIP holds, then
accurate reconstruction can be accomplished by solving the following
linear program:
min ||α̃||1 subject to: ΦΨα̃ = y.
α̃∈RN
Since our measurement vector y is unavoidably noisy (due to the
limited word length of the integer operation on the embedded node),
we rather solve the previous l1 minimization with relaxed constraints
for reconstruction given by:
min ||α̃||1 subject to: ||ΦΨα̃ − y||2 ≤ σ,
α̃∈RN
where σ bounds the amount of noise corrupting the data. Many
algorithms were introduced to solve this reconstruction problem,
including interior-point algorithms [12], gradient projection [13],
iterative thresholding [14], and greedy approaches such as orthogonal
matching pursuit (OMP) [15]. The results reported in this work are
based on the basis pursuit denoise algorithm provided in the SPGL1
solver [16].
III. ECG DATABASE AND P ERFORMANCE M ETRICS
To validate the performance of the considered compression scheme,
we use the MIT-BIH Arrhythmia Database [17], [18] that is the
most commonly used database for the comparative study of ECG
compression algorithms. This database contains 48 half-hour excerpts
of two-channel ambulatory ECG recordings, obtained from 47 sub-
jects studied by the BIH Arrhythmia Laboratory. The recordings
were digitized at 360 samples per second per channel with 11-
bit resolution over a 10 mV range. Moreover, to quantify the
compression performance while assessing the diagnostic quality of
the compressed ECG records, we employ the two most widely
used performance metrics, namely the compression ratio (CR) and
percentage root-mean-square difference (P RD). CR is defined as:
borig − bcomp
CR = × 100,
borig
where borig and bcomp represent the number of bits required for
the original and compressed signals, respectively. The P RD, and
associated signal-to-noise ratio (SN R), quantifies the percent error
between the original signal vector x and the reconstructed x̃:
||x − x̃||2
P RD = × 100, (5)
||x||2
SN R = −20 log10 (0.01P RD) . (6)
The link between the measured P RD and the diagnostic distortion
is established based on the work of Zigel et al. [19], which classifies
the different values of PRD based on the signal quality perceived by a
specialist. As it turns out, a reconstruction with P RD value between
0 and 2% corresponds to ”very good” quality, while a P RD value
between 2 and 9% is associated with ”good” quality. Finally, a P RD
H
value above 9% does not allow to determine the quality group.
IV. R EAL -T IME ECG C OMPRESSION
Our target platform is the Shimmer wireless sensor mote [2]. From
the hardware viewpoint, the ShimmerTM mainboard includes the low-
(1) power Texas Instrument 16-bit MSP430F1611 microcontroller and
a Bluetooth module. The MSP430 microcontroller runs at 8 MHz,
has 10 kB of RAM, 48 kB of Flash and includes a fast hardware
multiplier, but does not include a floating-point unit. From the
software viewpoint, we used the open-source GCC 3.2.3 tool chain
for the MSP430 [20] to generate the binaries. Since our aim is to
validate CS-based compression on the standard MIT-BIH arrhythmia
database, we used the ShimmerTM serial port to read in the database
records re-sampled at 256 Hz, and to readout the encoded ECG data.
The implementation of Gaussian random sensing with matrix
(2) Φ ∈ RM ×N , requires the implementation of a Gaussian-distributed
random number generator on the embedded platform and the com-
putation of a large matrix multiplication. This is too complex, time
consuming, and not real-time for both the encoder and decoder. To
address this problem, we explored three different approaches to the
implementation of the random sensing matrix Φ: (1) We implemented
an 8-bit quantized version of a normal random number generator to
(3) form Φ; the normal random number generator on board still made
this approach not real-time. (2) We also tried to circumvent the on-
board generation of the normal random numbers by storing them on
the platform; the large dense matrix multiplication still remained a
main bottleneck. (3) We introduced an innovative approach to CS
implementation using a sparse binary sensing matrix Φ, i.e., sparse
binary sensing.
In a sparse binary√ matrix Φ, each column has exactly d nonzero
entries equal to 1/ d, with d  N . The positions of the d non-
zero elements are randomly chosen to keep the incoherence between
the columns of the sensing matrix. Such a sensing matrix satisfies a
different form of the RIP property of (1), so-called RIPp property.
An M × N matrix, Φ is said to satisfy RIPp , if for any S-sparse
vector α, we have:
(1 − δ) ||α||p ≤ ||ΦΨα||p ≤ (1 + δ) ||α||p , (7)
where the l2 norm is replaced by lp norm. It was proven in Theorem
2 of [21] that the RIP1 property of a sparse binary sensing matrix
with exactly d ones on each column suffices to guarantee a good
sparse approximation recovery by a linear program.
Obviously, the choice of the number of non-zero elements depends
on the sparsity of the signal. To find out the minimum value of d
that strikes the optimal trade-off between execution time and (signal)
(4) recovery/reconstruction error, sparse binary sensing matrices are
applied to all the records of the MIT-BIH Arrhythmia ECG database,
and the output SN R of the reconstructed signals is measured.
Figure 2 reports the resulting average output SN R vs. the number of
1745
 50

20 CR = 50 45 CS (before)

Output PRD (averaged over all Data)

CR = 55 CS (after)
Output SNR (averaged over all Data)

18 40 DWT (before)
16 CR = 60 35 DWT (after)

14 30
CR = 65
12
25
10 CR = 70
20
8 CR = 75
15
6
CR = 80 10
4
5
2
0
0 0 20 40 60 80 100
5 10 15 20
Compression Ratio (CR)
Number of nonzero elements in sensing matrix Φ
Fig. 2. Output SN R vs. d for different compression ratios (CR)
Fig. 4. Output P RD vs. CR for CS and DWT before and after inter-packet
redundancy removal and Huffman coding
22
Sparse sensing
cardiology, we limited our review to those state-of-the-art algorithms
Output SNR (averaged over all Data)

20 Gaussian sensing
that were shown to achieve competitive compression ratios while be-
18
ing amenable to efficient, fast and low-memory-footprint implemen-
16
tation on our target platforms. Interestingly, these algorithms are all
14 based on the digital wavelet transform (DWT), namely, the embedded
12 zerotree wavelet (EZW) [22], the set partitioning in hierarchical trees
10 (SPIHT) [23] and thresholding-based algorithms [24]. Furthermore,
8 the latter thresholding-based algorithm by Benzid et al. [24], where
6
a fixed percentage of wavelet coefficients are zeroed, was shown
to outperform EZW and SPIHT at a lower computational cost.
4
50 55 60 65 70 75 80 Therefore, we will herein use it as a baseline algorithm to benchmark
Compression Ratio (CR) the performance, memory footprint, computational complexity and
Fig. 3. Performance comparison between Gaussian and sparse sensing energy consumption of our CS-based ECG compression scheme.
non-zero elements d in the sparse binary sensing matrix Φ. Clearly, B. Reconstruction quality
the output SN R saturates after d = 12 non-zero elements, which is
Figure 4 compares the output P RD, averaged over all database
the value retained for the rest of our hardware implementation. Note
records, for CS and DWT-based compression before and after inter-
that, in all our experiments, the sparse recovery problem of (3) was
packet redundancy removal and Huffman coding for different com-
solved using the SPGL1 solver [16] in Matlab c
.
pression ratios. It confirms the crucial role of the redundancy removal
Figure 3 shows the average output SN R vs. different compression
module and the careful design of the Huffman encoding. On average,
ratios (CR) for the two different approaches to the implementation
”very good” quality of reconstruction for DWT-based compression
of the random sensing matrix Φ explored in this section, namely,
is achieved up to CR = 73%, and up to CR = 51% for CS.
(1) Gaussian random sensing: the double-precision Matlab version;
Moreover, ”good” reconstruction quality is reached on average for
(2) Sparse binary sensing:
√ the version with d = 12 and all non-zero
up to CR = 90% and CR = 71% for DWT and CS, respectively.
entries equal to 1/ 12 implemented on the MSP430. These results
were obtained for an input vector of N = 512 samples and a 12- C. Memory usage
bit resolution for the input vector x and the measurement vector y.
The complete implementation of the thresholding-based DWT
Interestingly, the obtained results validate that there is no meaningful
compression requires 4.6 kB of RAM memory and 10 kB of Flash
performance difference between these two approaches.
memory, 5 kB of which is used for Huffman codebook storage. For
The use of a fixed binary sensing matrix, combined with the quasi-
CS, the complete implementation requires 6.5 kB of RAM and 7.5
periodic nature of the ECG signal, yields very similar consecutive
kB of Flash, 1.5 kB of which are for Huffman codebook storage.
measurement vectors y. So, there is a large inter-packet redundancy
that must be removed prior to encoding and wireless transmission. D. Power and energy consumption
Exploiting this fact, the redundancy removal module in Figure 1
To measure the power consumption during operation, we place
computes the difference between consecutive vectors, and only this
a 10.3Ω resistor in the power path of the platform. The voltage is
difference is further quantized. At the end, a state-of-the-art entropy
then readily measured using an oscilloscope, and the corresponding
coding module is used for further compression. Since the range of
current and power consumption are calculated. Our experimental
the difference signal just before encoding was found to be between
set-up consists of a Shimmer node implementing embedded ECG
[−256 : 255], a Huffman codebook of size 512 is needed with a
compression and wirelessly transmitting the compressed data using
maximum codeword length of 16 bits, for a given compression ratio.
its IEEE 802.15.4-compliant radio to a remote base station (a desktop
V. P ERFORMANCE A SSESSMENT AND B ENCHMARKING PC to which a second Shimmer node is connected). For simplicity,
a simple medium access control (MAC) scheme is implemented
A. Baseline thresholding-based DWT Compression between the node and the base station in FreeRTOS: the base station
Since our focus is on ECG compression techniques implementable periodically sends a beacon to maintain node synchronization, and the
in real-time on WBSN nodes for future ”personal” or ”wearable” tele- ECG sensing node sends its data in its guaranteed time slots (GTS),

1746
 TABLE I
N ODE LIFE TIME FOR ” GOOD ” RECONSTRUCTION QUALITY WITH
EMBEDDED COMPRESSION
DWT CS No Comp.
Compression Ratio (%) 90 71 0
Code execution time (ms) 580 99.5 0 Project BioCS-Node, which is financed by the Swiss Confederation.
Packet Ready every ... (ms) 2968.8 1023.7 296.9
Beacon Interval (ms) 11875 4094 1187
Energy Consumption (mJ) 8.78 7.46 8.37
Life time (h) (280 mAh@3.7 V ) 113.76 133.95 119.4
which is assigned by the base station and notified in the beacon. This
simple MAC is in fact compatible with the beacon-enabled mode of
the IEEE 802.15.4 MAC protocol [25].
To compress an ECG data vector of 512 samples (corresponding
to a duration of 2 seconds at sampling rate of 256 Hz), the
embedded CS code executes in 99.5 ms, whereas the DWT-based
code requires 580 ms. This confirms the relevance of CS as a
low-complexity encoding scheme. To study the corresponding node
lifetime, power consumption of the different working modes of the
node was characterized. Our measurements show an average node
power consumption of 84.6 mW during radio reception, and of
64.35 mW during transmission. The radio is only turned on for
beacon reception (i.e., 2.7 ms), and when there is a data packet
to be transmitted. Moreover, each transmitted packet is 127-Byte
long with 13 Byte of MAC overhead (4.2 ms). The rest of time,
the radio is off and the processor is idle and only wakes up
for sampling to preserve energy, which corresponds to an average
power consumption of 6.44 mW . Accordingly, Table I compares
the energy consumption and the resulting node lifetime (calculated
for a 280 mAh battery at 3.7 V ) of the two algorithms for a
”good” reconstruction quality. It also includes the same figures for the
streaming of uncompressed ECG. Interestingly, the reported results
show that CS achieves a relative node lifetime extension of 17.7 %
over DWT-based compression. This confirms the superior energy
efficiency of CS for a given reconstructed signal quality. This superior
energy efficiency is achieved thanks to CS’s lower complexity and
shorter processing time. Furthermore, the results of Table I establish
that DWT-based compression actually fails to achieve any lifetime
extension, compared to the default alternative of streaming uncom-
pressed ECG data. Alternatively, CS shows a 12.9 % extension in
node lifetime. In fact, these promising results could have been even
more impressive had the MSP430 been optimized for ultra-low-power
DSP operations. Since this generic low-power microcontroller is not,
it effectively limits the achievable node lifetime extension through
embedded signal pre-processing prior to wireless communication.
VI. C ONCLUSION
This paper proposed a complete system-level comparison between
a new CS-based and the state-of-the-art DWT-based embedded ECG
compression algorithms. As expected, nonadaptive CS-based com-
pression was found to exhibit inferior compression performance
compared to its signal-adaptive DWT-based counterpart for a given
reconstructed signal quality. The presented results, however, were
obtained using the default basis pursuit denoise algorithm for sparse
reconstruction, where no attempt has been made to exploit the highly
structured nature of the ECG signal. Despite its inferior compression
performance, CS-based compression was found to exhibit the best
overall energy efficiency thanks to its lower complexity and reduced
CPU execution time for ShimmerTM WBSN platform. As such,
our results validate the suitability of compressed sensing for real-
time energy-aware ECG compression on resource-constrained WBSN
1747
motes.
VII. ACKNOWLEDGEMENT
This research has been partially funded by the Nano-Tera.ch NTF
R EFERENCES
[1] Toumaz Technology, 2009. [Online]. Available: http://www.toumaz.
com/public/news.php?id=92
[2] Shimmer Research. [Online]. Available: http://shimmer-research.com
[3] R. F. Yazicioglu and et al., “Ultra-low-power wearable biopotential
sensor nodes,” in Proc. of the IEEE EMBC’09, Sep. 2009.
[4] L. Sörnmo and P. Laguna, Bioelectrical Signal Processing in Cardiac
and Neurological Applications. Elsevier Academic Press, 2005.
[5] E. Candès, J. Romberg, and T. Tao, “Robust uncertainty principles: Exact
signal reconstruction from highly incomplete frequency information,”
IEEE Trans. on Inform. Theory, vol. 52, no. 2, pp. 489–509, Feb. 2006.
[6] D. L. Donoho, “Compressed sensing,” IEEE Trans. on Inform. Theory,
vol. 52, no. 4, pp. 1289–1306, Apr. 2006.
[7] E. Candès and T. Tao, “Near optimal signal recovery from random
projections: Universal encoding strategies,” IEEE Trans. on Inform.
Theory, vol. 52, no. 12, pp. 5406–5425, Dec. 2006.
[8] S. Aviyente, “Compressed sensing framework for EEG compression,” in
Proceedings of the IEEE Workshop on Stat. Signal Proc. (SSP’07), Aug.
2007, pp. 181–184.
[9] S. Şenay, L. F. Chaparro, M. Sun, and R. J. Sclabassi, “Compressive
sensing and random filtering of EEG signals using slepian basis,” in
Proceedings of the EURASIP Signal Proc. Conf. (EUSIPCO’08), Aug.
2008.
[10] A. M. Abdulghani, A. J. Casson, and E. Rodriguez-Villegas, “Quantify-
ing the feasibility of compressive sensing in portable electroencephalog-
raphy systems,” in Proceedings of the Intl. Conf. on Foundations of
Augmented Cognition, Neuroergonomics and Operational Neuroscience
(FAC’09), 2009, pp. 319–328.
[11] E. J. Candès and T. Tao, “Decoding by linear programming,” IEEE
Trans. on Inform. Theory, vol. 51, no. 12, pp. 4203–4215, Dec. 2005.
[12] E. van den Berg and et al., “Probing the pareto frontier for basis pursuit
solutions,” SIAM J. on Sci. Computing, vol. 31, no. 2, 2008.
[13] M. Figueiredo and et al., “Gradient projection for sparse reconstruction:
Application to compressed sensing and other inverse problems,” IEEE
Jrl. of Selec. topics in Signal Proc., vol. 1, no. 4, pp. 586–597, 2007.
[14] I. Daubechies, M. Defrise, and C. D. Mol, “An iterative thresholding
algorithm for linear inverse problems with a sparsity constraint,” Comm.
Pure Appl. Math., vol. 57, pp. 1413–1457, 2004.
[15] J. Tropp, “Greed is good: Algorithmic results for sparse approximation,”
IEEE Trans. on Inform. Theory, vol. 50, no. 10, pp. 2231–2242, 2004.
[16] E. v. Berg and M. P. Friedlander, “SPGL1: A solver for large-
scale sparse reconstruction,” Jun. 2007. [Online]. Available: http:
//www.cs.ubc.ca/labs/scl/spgl1
[17] G. B. Moody and R. G. Mark, “The impact of the MIT-BIH arrhythmia
database,” IEEE Eng. in Med. and Bio., vol. 20, no. 3, pp. 45–50, 2001.
[18] “MIT-BIH arrhythmia database.” [Online]. Available: http://www.
physionet.org/physiobank/database/mitdb/
[19] Y. Zigel, A. Cohen, and A. Katz, “The weighted diagnostic distortion
(WDD) measure for ECG signal compression,” IEEE Trans. on Biomed.
Eng., vol. 47, pp. 1422–1430, 2000.
[20] “The GCC toolchain for the Texas Instruments MSP430 MCUs.”
[Online]. Available: http://mspgcc.sourceforge.net/
[21] R. Berinde and et al., “Combining geometry and combinatorics: A
unified approach to sparse signal recovery,” in Proc. of Allerton, 2008.
[22] M. L. Hilton, “Wavelet and wavelet packets compression of electrocar-
diogram,” IEEE Trans. on Biomed. Eng., vol. 44, no. 5, pp. 394–402,
1997.
[23] Z. Lu, D. Y. Kim, and W. A. Pearlman, “Wavelet compression of ECG
signals by the set partitioning in hierarchical trees algorithm,” IEEE
Trans. on Biomed. Eng., vol. 47, no. 7, pp. 849–856, 2000.
[24] R. Benzid, F. Marir, A. Boussaad, M. Benyoucef, and D. Arar, “Fixed
percentage of wavelet coefficients to be zeroed for ECG compression,”
Electronics Letters, vol. 39, no. 11, pp. 830–831, 2003.
[25] “Wireless medium access control (MAC) and physical layer (PHY) spec-
ifications for low-rate wireless personal area networks (LR-WPANs),”
IEEE, Tech. Rep. IEEE 802.15.4 Std., 2003.