Professional Documents
Culture Documents
Paper Ecg Cs Codec
Paper Ecg Cs Codec
Abstract—Remote monitoring of electrocardiogram (ECG) ECG signal compression has been an active area of interest
signals plays a critical role in the management of cardiovascular for several decades. Extensive surveys can be found in [6],
diseases. Current long term ECG monitoring systems generate [7]. Compressive sensing (CS) based techniques for ECG data
a large amount of data that need to be sent across a wireless
body area network. The wearable devices are resource limited. compression have been reviewed in [8], [9].
Hence a low resource consuming compression scheme is desirable. Transform domain techniques (e.g., Discrete Cosine Trans-
Compressive sensing (CS) is quite appealing as a low complexity form [10], Discrete Cosine Transform [11], [12], Discrete
method for compression of ECG data. This paper investigates Wavelet Transform [13]–[16]) are popular in ECG compres-
the distribution of the compressive measurements and proposes sion and achieve high compression ratios (CR) at clinically
a CS based codec architecture comprising simple quantization
and asymmetric numeral systems (ANS) based entropy coding acceptable quality. However, they require computationally
steps that can significantly boost the compression ratio without intensive sparsifying transforms on all data samples and are
sacrificing reconstruction performance. The encoder has low thus not suitable for WBAN sensor nodes [8].
computational complexity beyond the sensing operation and can Compressive sensing [17]–[21] uses a sub-Nyquist sampling
be implemented entirely using integer arithmetic. The quantized method by acquiring a small number of incoherent measure-
Gaussian entropy model for the compressive measurements is
estimated directly from the data and can be easily adapted ments which are sufficient to reconstruct the signal if the signal
to achieve better compression. Our decoder architecture is is sufficiently sparse in some basis. For a sparse signal x ∈ Rn ,
flexible in choosing any suitable reconstruction method. We have one would make m linear measurements where m ≪ n which
tested our encoder with bound optimization based block sparse can be mathematically represented by a sensing operation
Bayesian learning (BSBL-BO) reconstruction algorithm on MIT-
BIH Arrhythmia database. We have validated that our encoder y = Φx (1)
can achieve additional 20-30% of space savings without any
impact on reconstruction quality on top of the savings in terms where Φ ∈ Rm×n is a matrix representation of the sensing
of measurements. We have also clearly described the digitization process and y ∈ Rm the set of m measurements collected for
process for the measurements which has often been ignored
in the literature on CS based compression of ECG data. As
x. A suitable reconstruction algorithm can recovery x from y.
part of reproducible research we have open sourced our codec Ideally, the sensing process should be implemented at the
(implemented in Python with JAX based GPU acceleration for hardware level in the analog to digital conversion (ADC)
the decoder). process. However, much of the use of CS in ECG follows a
digital CS paradigm [5] where the ECG samples are acquired
first by the ADC circuit on the device and then they are
I. I NTRODUCTION
translated into incoherent measurements via the multiplication
Noncommunicable diseases (NCDs) account for 72% of all of a digital sensing matrix. These measurements are then
global deaths with cardiovascular diseases (CVDs) accounting transmitted to remote telemonitoring servers. A suitable recon-
for 44% of all NCD mortality [1]. For patients with CVDs, struction algorithm is used on the server to recover the original
wearable device based remote ECG monitoring plays a critical ECG signal from the compressive measurements. Reconstruc-
role in their disease management and care. tion algorithms for ECG signals include: greedy algorithms
In wireless body area networks (WBAN) based telemoni- [22] (simultaneous orthogonal matching pursuit), optimization
toring networks [2], the energy consumption on sensor nodes based algorithms [23], [5] (SPG-L1), Bayesian learning based
is a primary design constraint [3]. The wearable sensor nodes algorithms [24]–[26], deep learning based algorithms [27].
are often battery operated. It is necessary to reduce energy In order to keep the sensing matrix multiplication simple
consumption as much as possible. and efficient, sparse binary sensing matrices are a popular
However, long term ECG monitoring can generate a large choice [5], [24].
amount of uncompressed data. For example, each half hour
2 lead recording in the MIT-BIH Arrhythmia database [4]
requires 1.9MB of storage. As shown in [5], in a real time A. Problem Formulation
telemonitoring sensor node, the wireless transmission of data We consider the problem of efficient transmission of com-
consumes most of the energy. The real time compression of pressive measurements of ECG signals over the wireless body
ECG data by a low complexity encoder has received significant area networks under the digital compressive sensing paradigm.
attention in the past decade. Let x be an ECG signal and y be the corresponding stream
2
of compressive measurements. Our goal is to transform y into lossless data compression. zstd by Facebook Inc. [34] and
a bitstream s with as few bits as possible without losing the LZFSE by Apple Inc. are popular lossless data compression
signal reconstruction quality. formats based on ANS. To the best of our knowledge, ANS
We consider whether a digital quantization of the compres- type stream codes have not been considered in the past for the
sive measurements affects the reconstruction quality. Further, entropy coding of compressive measurements.
we study the empirical distribution of compressive measure-
ments to explore efficient ways of entropy coding of the C. Contributions
quantized measurements. In this paper, we present an adaptive quantization and
A primary constraint in our design is that the encoder entropy coding scheme for the digital compressive measure-
should avoid any floating point arithmetic so that it can be ments in the encoders running on the WBAN sensor nodes.
implemented efficiently in low power devices. The adaptive quantization scheme has been designed in a
manner so that the quantization noise is kept within a specified
B. Related Work threshold. Once the measurements are quantized and clipped,
they are ready to be entropy coded. A quantized Gaussian
The literature on the use of CS for ECG compression is
distribution of the measurements is estimated directly from
mostly focused on the design of the specific sensing matrix,
the data and the parameters for this distribution are used for
sparsifying dictionary, or reconstruction algorithm for the high
the entropy coding of the measurements. Asymmetric numeral
quality reconstruction of the ECG signal from the compressive
systems (ANS) based entropy coding is then used for efficient
measurements. To the best of our knowledge, (digital) quan-
entropy coding of the quantized and clipped measurements.
tization and entropy coding of the compressive measurements
The ECG signal is split into frames for encoding purposes.
of ECG data hasn’t received much attention in the past.
We have designed a bitstream format with a stream header to
Mamaghanian et al. [5] use a Huffman codebook which is
describe the encoding parameters and a frame header for each
deployed inside the sensor device. They don’t use any quan-
encoded frame containing frame wise quantization parameters.
tization of the measurements. However, they don’t provide
We have designed a decoder which accepts this encoded
much detail on how the codebook was designed or how should
bitstream and reconstructs the ECG signal frame by frame
it be adapted for variations in ECG signals. They clearly
in real time.
define the compression ratio in terms of a ratio between the
Our encoding scheme doesn’t require a fixed codebook for
uncompressed bits bitsu and the compressed bits bitsc . They
u −bitsc
entropy coding. The encoder has been designed in a manner
define it as bitsbits × 100. This is often known in literature
u so that all the operations can be implemented by integer arith-
as percentage space savings.
metic. We have tested our encoder on MIT-BIH Arrhythmia
Simulation based studies often send the real valued com-
database and are able to demonstrate additional space savings
pressive measurements to the decoder modules and don’t
beyond the savings by reduced number of measurements. Our
consider the issue of number of bits required to encode each
implementation is available as part of reproducible research
measurement. Zhang et al. [24], Zhang et al. [27] define
n−m on GitHub [35].
n × 100 as compression ratio. Mangia et al. [28] define
n
m as the compression ratio. Polania et al. [29] define m n
D. Paper Organization
as compression ratio. Picariello et al. [30] use a non-random
sensing matrix. They infer a circulant binary sensing matrix The rest of the paper is organized as follows. Section II
directly from the ECG signal being compressed. The sensing describes the ECG database used for the evaluation of the
matrix is adapted as and when there is significant change in codec and the performance metrics. Section III describes the
the signal statistics. They represent both the ECG samples proposed codec architecture. The analysis of the experimental
and compressive measurements with same number of bits results is covered in section IV. Section V summarizes our
per sample/measurement. However, in their case, there is the contributions and identifies future work. Appendix A provides
additional overhead of sending the sensing matrix updates. a short overview on compressive sensing. Appendix C provides
n a short overview on entropy coding.
Their compression ratio is slightly lower than m where they
n
call m as the under-sampling ratio.
Huffman codes have frequently been used in ECG data II. ECG DATABASE AND P ERFORMANCE M ETRICS
compression for non-CS methods. Luo et al. in [31] proposed We use the MIT-BIH Arrhythmia Database [4] from Phy-
a dynamic compression scheme for ECG signals which con- sioNet [36]. The database contains 48 half hour excerpts of
sisted of a digital integrate and fire sampler followed by an two channel ambulatory ECG recordings from 47 subjects.
entropy coder. They used Huffman codes for entropy coding of The recordings were digitized at 360 samples per second for
the timestamps associated with the impulse train. Chouakri et both channels with 11 bit resolution over a 10mV range.
al. in [32] compute the DWT of ECG signal by Db6 wavelet, The samples can be read in both digital (integer) form or
select coefficients using higher order statistics thresholding, as physical values (floating point) via the software provided
then perform linear prediction coding and Huffman coding of by PhysioNet. We use the integer values in our experiments
the selected coefficients. since our encoder is designed for integer arithmetic. We use
Asymmetric Numeral Systems (ANS) [33] based entropy the MLII signal (first channel) from each recording in our
coding schemes have seen much success in recent years for experiments.
3
together to form the (signed) signal matrix 4 : 6) Adaptive Clipping: We clip the values in ȳ to the range
[µy − rσy , µy + rσy ] where r is the range parameter estimate
X = x1 x2 . . . xw . (12)
from the data. This is denoted as
2) Compressive Sensing: The digital compressive sensing ŷ = clip(ȳ, µy − rσy , µy + rσy ). (18)
is represented as
yi = Φxi (13) Similar to adaptive quantization, we vary r from 2 to 8 till we
have captured sufficient variation in ȳ and N RMSE(y, ŷ) ≤
where yi is the measurement vector for the i-th window. γ.
Combining the measurement vectors for each window, the 7) Entropy Coding: We then model the measurement values
measurement matrix for the entire record is given by in ŷ as a quantized Gaussian distributed random variable with
Y = ΦX. (14) mean µy , standard deviation σy , minimum value µy −rσy and
maximum value µy + rσy . We use the ANS entropy coder to
Note that by design, the sensing operation can be implemented encode ŷ into an array c of 32-bit integers (called words).
using just lookup and integer addition. The ones in each row This becomes the payload of the frame to be sent to the
of Φ identify the samples within the window to be picked up decoder. The total number of compressed bits in the frame
and summed. Consequently, Y consists of only signed integer payload is the length of the array nc times 32. Note that we
values. have encoded and transmitted ŷ and not the unclipped ȳ. ANS
3) Flattening: Beyond this point, the window structure of entropy coding is a lossless encoding scheme. Hence, ŷ will be
signal is not relevant for quantization and entropy coding reproduced faithfully in the decoder if there are no bit errors
purposes in our design. Hence, we flatten it (column by involved in the transmission 5 .
column) into a vector y of mw measurements. 8) Integer Arithmetic: The input to digital compressive
y = flatten(Y). (15) sensing is a stream of integers. The sensing process with
sparse binary sensing matrix can be implemented using integer
4) Quantization: Next, we perform a simple quantization of sums and lookup. It is possible to implement the computation
measurement values. If fixed quantization has been specified of approximate mean and standard deviation using integer
in the stream header, then for each entry y in y, we compute arithmetic. We can use the normalized mean square error based
thresholds for adaptive quantization and clipping steps. ANS
ȳ = y/2q . (16)
entropy coding is fully implemented using integer arithmetic.
For the whole measurement vector, we can write this as Thus, the proposed encoder can be fully implemented using
1 integer arithmetic.
ȳ = q y. (17)
2
This can be easily implemented in a computer as a signed right C. Decoding
shift by q bits (for integer values). This reduces the range of The decoder initializes itself by reading the stream header
values in y by a factor of 2q . and creating the sensing matrix to be used for the decoding
If adaptive quantization has been specified, then we vary the of compressive measurements frame by frame.
quantization parameter q from a value qmax = 6 down to a
value qmin = 0. For each value of q, we compute ȳ = 21q y. We Compressed
Bits
Entropy
Decoding
Inverse
Quantization
Windowing
then compute ỹ = 2q ȳ. We stop if N RMSE(y, ỹ) reaches a
limit specified by the parameter ρ in the stream header.
Reconstruction Decoded
5) Entropy Model: Before we explain the clipping step, we Algorithm
Flattening
Samples
shall describe our entropy model. We model the measurements
as samples from a quantized Gaussian distribution which can Fig. 2. Digital Compressive Sensing Decoder
only take integral values. First, we estimate the mean µy and
standard deviation σy of measurement values in ȳ. We shall Here we describe the decoding process for each frame.
need to transmit µy and σy in the frame header. We round the Figure 2 shows a block diagram for the decoding process.
values of µy and σy to the nearest integer so that it can be Decoding of a frame starts by reading the frame header which
transmitted efficiently as integers. provides the frame encoding parameters: µy , σy , q, r, nw , nc .
Entropy coding works with a finite alphabet. Accordingly, 1) Entropy Decoding: The frame header is used for build-
the quantized Gaussian model requires specification of the ing the quantized Gaussian distribution model for the decoding
minimum and maximum values that our quantized measure- of the entropy coded measurements from the frame payload.
ments can take. The range of values in ȳ must be clipped to The minimum and maximum values for the model are com-
this range. puted as:
4 PhysioNet provides the baseline values for each channel in their ECG ymin , ymax = µy − rσy , µy + rσy . (19)
records. Since the digital samples are unsigned, we have subtracted them by
the baseline value (1024 for 11-bit encoding). 11 bits mean that unsigned
nc tells us the number of words (4nc bytes) to be read from
values range from 0 to 2047. The baseline for zero amplitude is digitally the bitstream for the frame payload. The ANS decoder is used
represented as 1024. After baseline adjustment, the range of values becomes
[−1024, 1023]. 5 We assume that appropriate channel coding mechanism has been used.
6
(b) shows that PSS (percentage space saving) increases V. C ONCLUSION AND F UTURE W ORK
linearly with q. As q increases, the size of the alphabet for
A. Contributions
entropy coding reduces and this leads to increased space
savings. The key contributions of our encoder architecture are the
(c) shows the variation of the quality score with q and we following:
can clearly see that quality score generally increases till q = 4 • We have presented a digital compressive sensing based
after which it starts decreasing. encoder design including adaptive digital quantization
The key result is shown in panel (d) which shows the and entropy coding which can be implemented entirely
variation of PSS with PMS at different values of q. PSS using integer arithmetic.
increases linearly with PMS. Also, PSS is much higher than • We have demonstrated that ANS based arithmetic coding
PMS at higher quantization levels. can lead to significant space savings.
We have also listed the numerical values in table IV. The • We have demonstrated that we don’t need a hard-coded
first column for q = 0 depicts the case where no quantization code-book for entropy coding. We can simply model the
has been applied, we are only doing clipping and entropy measurement values distributed as quantized Gaussian
coding. Since PSS is consistently higher than PMS, it is clear values for effective entropy coding.
that we need less than 11 bits per sample on an average • We use simple mean and standard deviation as the entropy
after entropy coding of the measurement values. Increasing model parameters which can be dynamically adapted. The
quantization linearly increases the PSS. Since up to q = 4, overhead of the side information is negligible.
there is no noticeable impact on PRD, as seen in panel (a), it • We have demonstrated up to 16% of additional space sav-
is safe to enjoy these savings in bit rate. At 40% PMS, one ings contributed by the quantization and entropy coding
can attain up to 69.3% PSS without losing any reconstruction steps.
quality while one can push further up to q = 6 to the PSS • We have described a complete bitstream specification for
of 79.9% with some degradation in quality while still staying carriage of all the side information needed to decode the
under acceptable PRD. entropy coded compressive measurements.
The software code implementing this codec and all scripts
TABLE IV for the experimental studies conducted in this work have been
PSS AT DIFFERENT VALUES OF q AND PMS FOR RECORD 100 released as open source software on GitHub [35].
q 0 1 2 3 4 5 6 7
PMS
25.0
MR
0.2
22.5 90 70 0.25
0.3
20.0 0.35
0.4
80 60 0.45
17.5 0.5
0.6
15.0
PSS (%)
PRD %
70 50
QS
12.5
MR MR
0.2 0.2
10.0 0.25 0.25 40
0.3 60 0.3
7.5 0.35 0.35
0.4 0.4
0.45 0.45 30
5.0 0.5 50 0.5
0.6 0.6
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
q q q
6
MR MR
0.2 10 0.2
90 0.25 0.25
5 0.3 0.3
0.35 9 0.35
0.4 0.4
80 0.45 0.45
0.5 8 0.5
4
0.6 0.6
7
PSS (%)
bpm
bps
70
3
q 6
0
60 1 5
2 2
3
4 4
5
50 6 1
7 3
40 45 50 55 60 65 70 75 80 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
PMS (%) q q
TABLE V TABLE VI
M EASUREMENT STATISTICS AT m = 256, n = 512, d = 4 C OMPRESSION STATISTICS AT m = 256, n = 512, d = 4
using incoherent measurements. The basic model can be linear measurements. Each row of Φ represents one linear
expressed as: measurement as a linear functional on x. y is a measurement
y = Φx + e (25) vector consisting of m distinct measurements done on x. By
design Φ is a full rank matrix. Hence every set of m columns
where x ∈ Rn is a signal to compress (of length n). of Φ is linearly independent. e ∈ Rm is the error/noise
Φ ∈ Rm×n is a sensing matrix which compresses x via introduced during the measurement process. In our digital CS
10
[9] S. S. Kumar and P. Ramachandran, “Review on compressive sensing [32] S. A. Chouakri, O. Djaafri, and A. Taleb-Ahmed, “Wavelet transform
algorithms for ecg signal for iot based deep learning framework,” and huffman coding based electrocardiogram compression algorithm:
Applied Sciences, vol. 12, no. 16, p. 8368, 2022. Application to telecardiology,” in Journal of Physics: Conference Series,
[10] H. Al-Nashash, “A dynamic fourier series for the compression of ecg vol. 454, no. 1. IOP Publishing, 2013, p. 012086.
using fft and adaptive coefficient estimation,” Medical engineering & [33] J. Duda, “Asymmetric numeral systems: entropy coding combining
physics, vol. 17, no. 3, pp. 197–203, 1995. speed of huffman coding with compression rate of arithmetic coding,”
[11] L. V. Batista, E. U. K. Melcher, and L. C. Carvalho, “Compression arXiv preprint arXiv:1311.2540, 2013.
of ecg signals by optimized quantization of discrete cosine transform [34] Collet, Yann, “Zstandard,” 2022. [Online]. Available: https://facebook.
coefficients,” Medical engineering & physics, vol. 23, no. 2, pp. 127– github.io/zstd
134, 2001. [35] Kumar, Shailesh, “Compressive sensing based codec for ecg,” 2022.
[12] A. Bendifallah, R. Benzid, and M. Boulemden, “Improved ecg com- [Online]. Available: https://github.com/shailesh1729/ecg
pression method using discrete cosine transform,” Electronics letters, [36] A. L. Goldberger, L. A. Amaral, L. Glass, J. M. Hausdorff, P. C.
vol. 47, no. 2, p. 1, 2011. Ivanov, R. G. Mark, J. E. Mietus, G. B. Moody, C.-K. Peng, and H. E.
[13] A. Djohan, T. Q. Nguyen, and W. J. Tompkins, “Ecg compression Stanley, “Physiobank, physiotoolkit, and physionet: components of a
using discrete symmetric wavelet transform,” in Proceedings of 17th new research resource for complex physiologic signals,” circulation, vol.
International Conference of the Engineering in Medicine and Biology 101, no. 23, pp. e215–e220, 2000.
Society, vol. 1. IEEE, 1995, pp. 167–168. [37] Z. Zhang, S. Wei, D. Wei, L. Li, F. Liu, and C. Liu, “Comparison
[14] Z. Lu, D. Y. Kim, and W. A. Pearlman, “Wavelet compression of ecg of four recovery algorithms used in compressed sensing for ecg signal
signals by the set partitioning in hierarchical trees algorithm,” IEEE processing,” in 2016 Computing in Cardiology Conference (CinC).
transactions on Biomedical Engineering, vol. 47, no. 7, pp. 849–856, IEEE, 2016, pp. 401–404.
2000. [38] Y. Zigel, A. Cohen, and A. Katz, “The weighted diagnostic distortion
[15] M. Pooyan, A. Taheri, M. Moazami-Goudarzi, and I. Saboori, “Wavelet (wdd) measure for ecg signal compression,” IEEE transactions on
compression of ecg signals using spiht algorithm,” International Journal biomedical engineering, vol. 47, no. 11, pp. 1422–1430, 2000.
of signal processing, vol. 1, no. 3, p. 4, 2004. [39] S. Kumar, “Cr-sparse: Hardware accelerated functional algorithms for
[16] B. S. Kim, S. K. Yoo, and M.-H. Lee, “Wavelet-based low-delay sparse signal processing in python using jax,” Journal of Open Source
ecg compression algorithm for continuous ecg transmission,” IEEE Software, vol. 6, no. 68, p. 3917, 2021.
Transactions on Information Technology in Biomedicine, vol. 10, no. 1, [40] R. Bamler, “Understanding entropy coding with asymmetric nu-
pp. 77–83, 2006. meral systems (ans): a statistician’s perspective,” arXiv preprint
[17] D. L. Donoho, “Compressed sensing,” Information Theory, IEEE Trans- arXiv:2201.01741, 2022.
actions on, vol. 52, no. 4, pp. 1289–1306, 2006. [41] Y. C. Eldar, P. Kuppinger, and H. Bolcskei, “Block-sparse signals:
[18] R. G. Baraniuk, “Compressive sensing [lecture notes],” Signal Process- Uncertainty relations and efficient recovery,” IEEE Transactions on
ing Magazine, IEEE, vol. 24, no. 4, pp. 118–121, 2007. Signal Processing, vol. 58, no. 6, pp. 3042–3054, 2010.
[19] E. J. Candès, “Compressive sampling,” in Proceedings of the Interna- [42] J. Rissanen and G. G. Langdon, “Arithmetic coding,” IBM Journal of
tional Congress of Mathematicians: Madrid, August 22-30, 2006: invited research and development, vol. 23, no. 2, pp. 149–162, 1979.
lectures, 2006, pp. 1433–1452. [43] I. H. Witten, R. M. Neal, and J. G. Cleary, “Arithmetic coding for data
[20] E. J. Candès and M. B. Wakin, “An introduction to compressive compression,” Communications of the ACM, vol. 30, no. 6, pp. 520–540,
sampling,” Signal Processing Magazine, IEEE, vol. 25, no. 2, pp. 21–30, 1987.
2008.
[21] E. J. Candes and T. Tao, “Near-optimal signal recovery from random
projections: Universal encoding strategies?” Information Theory, IEEE
Transactions on, vol. 52, no. 12, pp. 5406–5425, 2006.
[22] L. F. Polania, R. E. Carrillo, M. Blanco-Velasco, and K. E. Barner,
“Compressed sensing based method for ecg compression,” in 2011 IEEE
international conference on acoustics, speech and signal processing
(ICASSP). IEEE, 2011, pp. 761–764.
[23] J. Zhang, Z. Gu, Z. L. Yu, and Y. Li, “Energy-efficient ecg compression
on wireless biosensors via minimal coherence sensing and weighted l1
minimization reconstruction,” IEEE Journal of Biomedical and Health
Informatics, vol. 19, no. 2, pp. 520–528, 2014.
[24] Z. Zhang, T.-P. Jung, S. Makeig, and B. D. Rao, “Compressed sensing
for energy-efficient wireless telemonitoring of noninvasive fetal ecg
via block sparse bayesian learning,” IEEE Transactions on Biomedical
Engineering, vol. 60, no. 2, pp. 300–309, 2012.
[25] Z. Zhang, T.-P. Jung, S. Makeig, Z. Pi, and B. D. Rao, “Spatiotemporal
sparse bayesian learning with applications to compressed sensing of mul-
tichannel physiological signals,” IEEE transactions on neural systems
and rehabilitation engineering, vol. 22, no. 6, pp. 1186–1197, 2014.
[26] Z. Zhang and B. D. Rao, “Extension of sbl algorithms for the recovery
of block sparse signals with intra-block correlation,” IEEE Transactions
on Signal Processing, vol. 61, no. 8, pp. 2009–2015, 2013.
[27] H. Zhang, Z. Dong, Z. Wang, L. Guo, and Z. Wang, “Csnet: A
deep learning approach for ecg compressed sensing,” Biomedical Signal
Processing and Control, vol. 70, p. 103065, 2021.
[28] M. Mangia, L. Prono, A. Marchioni, F. Pareschi, R. Rovatti, and G. Setti,
“Deep neural oracles for short-window optimized compressed sensing
of biosignals,” IEEE transactions on biomedical circuits and systems,
vol. 14, no. 3, pp. 545–557, 2020.
[29] L. F. Polanı́a and R. I. Plaza, “Compressed sensing ecg using re-
stricted boltzmann machines,” Biomedical Signal Processing and Con-
trol, vol. 45, pp. 237–245, 2018.
[30] F. Picariello, G. Iadarola, E. Balestrieri, I. Tudosa, and L. De Vito,
“A novel compressive sampling method for ecg wearable measurement
systems,” Measurement, vol. 167, p. 108259, 2021.
[31] K. Luo, J. Li, and J. Wu, “A dynamic compression scheme for energy-
efficient real-time wireless electrocardiogram biosensors,” IEEE Trans-
actions on Instrumentation and Measurement, vol. 63, no. 9, pp. 2160–
2169, 2014.