M.Vithya, S.Meenakshi: Noise Cancellation in Fetal ECG Extraction Using SSNF Algorithm

International Journal of Applied Engineering Research, ISSN 0973-4562 Vol. 10 No.
46 (2015)
© Research India Publications; httpwww.ripublication.comijaer.htm
Noise Cancellation in Fetal ECG Extraction Using SSNF Algorithm

M.Vithya1 , S.Meenakshi 2
1 Dr. M .Suganthi and 2Mrs.S.Meenkashi
1. Student, M.E.,(Communication Systems), Mahendra College of Engineering,Salem-636106
1 Professor, Department of ECE, Mahendra College of Engineering, Salem - 636 106.
2.Assitant Professor, Department of ECE, Mahendra College of Engineering,Salem-636106
2 Assistant Professor, Department of ECE, Mahendra College of Engineering, Salem - 636 106.
Abstract
Extracting clean fetal electrocardiogram (ECG) signals is very important in fetal

monitoring. In this paper, we proposed a new method for fetal ECG extraction based on wavelet
analysis, the least mean square(LMS) adaptive filtering algorithm, and the spatially selective
noise filtration (SSNF) algorithm. First, abdominal signals and thoracic signals were processed
by stationary wavelet transform (SWT), and the wavelet coefficients at each scale were obtained.
For each scale, the detail coefficients were processed by the LMS algorithm. The coefficient of
the abdominal signal was taken as the original input of the LMS adaptive filtering system, and
the coefficient of the thoracic signal as the reference input. Then, correlations of the processed
wavelet coefficients were computed. The threshold was set and noise components were removed
with the SSNF algorithm. Finally, the processed wavelet coefficients were reconstructed by
inverse SWT to obtain fetal ECG. Twenty cases of simulated data and 12 cases of clinical
data were used. Experimental results showed that the proposed method outperforms the LMS
algorithm:
(1) it shows improvement in case of superposition R-peaks of fetal ECG and maternal ECG;
(2) noise disturbance is eliminated by incorporating the SSNF algorithm and the extracted
waveform is more stable; and
(3) the performance is proven quantitatively by SNR calculation. The results indicated that the
proposed algorithm can be used for extracting fetal ECG from abdominal signals.
32173
International Journal of Applied Engineering Research, ISSN 0973-4562 Vol. 10 No.46 (2015)
INTRODUCTION
Fetal electrocardiogram (ECG) waveform A number of techniques have been proposed

analysis is performed with the measurement of for fetal ECG extraction, such as adaptive filtering
electrical activity from fetal heart and has [4,5], blind source extraction (BSE) [6,7], singular
developed over the last 3 decades. It provides value decomposition (SVD) [8,9], independent
information about the physiological state of the fetus component analysis (ICA) [10–12] and neural
that can help clinicians to make appropriate and networks [13,14]. There are also some improved
timely decisions during labor. Fetal patho- logical methods [15–17] and combinations of different
characteristics can be detected by analysis of algorithms [18,19]. However, all these methods have
ECG waveform during the pregnancy period, and their own limitations. Adaptive filtering is simple
fetus mortality rate can be greatly reduced [1]. and fast, but fetal ECG extracted by this algorithm
Besides, compared with heart sound and heartbeat, still contains maternal ECG and other disturbance.
ECG varies more quickly and sensitively to the For ICA algorithm, statistical modeling of
abnorm- ality. Therefore, extraction of clean fetal probability density functions is challenging. Neural
ECG has vital significance for fetal monitoring [2]. networks have problems of bad generalization
capability, non-convergence, and selection of neuron
Fetal ECG is generally extracted from
functions or network structures. Therefore, it
maternal abdominal signals in the clinic (hereafter,
remains a difficult problem to extract clean fetal
we use “abdominal” for “maternal abdominal”, and
ECG from the signal recorded on abdomen.
“thoracic” for “maternal thoracic”). However, fetal
ECG signals are faint and mixed with several other In this paper, we propose a new technique
sources of disturbance [3]. In addition to maternal for fetal ECG extraction which combines adaptive
muscular noise and power line disturbance, the filtering algorithm with wavelet analysis. First,
strongest disturbance is maternal ECG which is 5 to abdominal signals and thoracic signals were pro-
10 times stronger than fetal ECG. Furthermore, cessed by wavelet transform, and the wavelet
much of fetal ECG coincides with maternal ECG coefficients at each scale were obtained. For each
both in time domain and frequency domain. scale, the detail coefficients were
Consequently, extracting fetal ECG from cutaneous
potential recording of a pregnant woman is a very
challenging task.
32174
processed by the adaptive filtering algorithm. where w(n) is the adaptive filter coefficient,
The coefficient of the abdominal signal was taken and μ is the chosen step size.
as the original input of the adaptive filtering
Wavelets is a mathematical tool which can
system, and the coefficient of the thoracic signal as
be used to extract information from a variety of
the reference input. Then, correlations of the
data types. Unlike the Fourier transform, wavelet
processed wavelet coefficients were computed. The
analysis gives a multi-resolution analysis of
threshold was set and noise components were
signals. It could focus on any signal′s details and
removed with the spatially selective noise filtration
is an efficient method in signal processing. Wavelet
algorithm. Finally, the processed wavelet
transform can decompose a signal into several
coefficients were reconstructed by inverse wavelet
scales that represent different frequency bands, and
transform to obtain fetal ECG.
at each scale, the positions of the signal′s instanta-
Methodology neous structures can be determined approximately.
Adaptive filtering algorithm When wavelet decomposition is performed

progressively by the discrete wavelet transform
As a classical algorithm, adaptive filtering
(DWT) [21], the length of the wavelet
has been used widely in fetal ECG extraction.
coefficients on each scale will become smaller.
The adaptive filtering configuration is shown in
Thus, DWT is not time-invariant and is sensitive to
Fig. 1. The adaptive filtering system has two inputs:
small shifts in the input signal. These shifts can lead
an original input and a reference input. The
to major variations and may also produce signals of
abdominal signal was used as the original input
jump discontinuity after reconstruction. The jump
d(n), and the thoracic signal as the reference input
discontinuity is called the Gibbs phenomenon. This
x(n). The coefficients of the adaptive filter were
problem can be avoided by removing down-
constantly adjusted with the feedback e(n) until the
sampling and up-sampling of the filter coefficients
output y(n) was very close to the maternal ECG
by a factor of 2j 1 in the jth scale in
component of the abdominal signal. Then, fetal
stationary wavelet transform (SWT). The output of
ECG was obtained by removing the maternal ECG
each scale in SWT contains the same number of
component from the abdominal signal.
samples as the input. Because the length of the SWT
The adaptive filter is adapted by the least coefficients at each scale is constant, SWT
mean square error (LMS) algorithm, which is a overcomes the lack of translation-invariance in
most widely used adaptive filtering algorithm. The DWT and can avoid the Gibbs phenomenon.
LMS update is given by Additionally, SWT can give a more approximate
32175
estimate of the continuous wavelet transform [22].
In this paper, SWT [23] was used.
Wavelet is a very powerful tool for
SWT employs two set of functions: the non-stationary signal analysis and has been
scaling function φðnÞ and the wavelet function widely used in fetal ECG extraction [24,25]. The
ΦðnÞ, which are defined on a chosen wavelet. limitation of this approach is that the extraction is
Using wavelet decomposition to process the signal greatly affected by abdominal signals. The wavelet
f(n), the approximation coefficient cj,k and the modulus maximum algorithm was used to extract
detail coefficient dj,k at the jth scale at the kth fetal ECG from abdominal signals [26], but
moment are obtained calculation of the alternate projection algorithm is
too complex for real-time applications [27,28].
The adaptive filtering algorithm has a
Generally, wavelet analysis algorithms are more
number of advantages over BSE and SVD, such as
often used to de-noise signals before or after fetal
its transient and steady-state performance,
ECG extraction.
computational simplicity in most situations, and
good real- time capability. The main limitation is Wavelet coefficient correlation algorithm
that some maternal ECGs would remain in the
Three kinds of wavelet de-noising methods
extracted fetal ECG [20].
[29] are widely used. They are wavelet
Wavelet analysis coefficients of modulus maximum [30], spatially
selective noise filtration (SSNF) [31], and wavelet
Wavelet analysis is a branch of applied
threshold de-noising [32]. SSNF uses the direct
mathematics that has produced a collection of tools
spatial correlation of wavelet transform at some
for signal and image processing
adjacent scales. Coefficients of the signal show
strong correlation over several wavelet scales, while
coefficients of the noise do not, so a spatial
correlation calculation operation can enhance the
signal and suppress the noise.
Fig. 1. Adaptive filtering configuration.
32176
Wavelet coefficients related to signals can be extracted by M 1 T

comparing the value of NCorr.
(4) The processed wavelet coefficients

were reconstructed by inverse SWT to obtain fetal
Fetal ECG Extraction
ECG.
Adaptive filtering and wavelet transform
Performance index of fetal ECG
are both classical signal processing algorithms. In
extraction
this paper, the LMS adaptive algorithm and the
wavelet coefficient correlation algorithm were The method of fetal ECG extraction can be
employed in the wavelet domain. evaluated by both observation and quantitative
performance. Extracted waveforms show elimination
Extraction of fetal ECG includes the
of maternal ECG directly, and R-peaks of fetal
following steps:
ECG are localized. R-peaks were detected by the
(1) The abdominal signal and the thoracic difference algorithm [34,35].
signal were processed by SWT, and then wavelet
Let f(n) be the original signal, and D(n) be
coefficients at each scale were obtained. In
the second-order difference of f(n),
wavelet decomposition, the bior1.5 wavelet from the
Matlab wavelet toolbox was selected after If the adjacent elements D(i) to D(k), (i o k)
comparison of various wavelet families at are all positive ones, the threshold Thr is defined as
different parameters, and the decomposition scale estimation of fetal ECG, and p(.) is the signal piece.
was set as 5 according to frequency
Experimental method
characteristics.
To verify the feasibility of the proposed
(2) At each scale, the detail coefficients
method, experiments were conducted using 20 cases
were processed by the LMS adaptive algorithm.
of simulated data and 12 cases of clinical data. All
The coefficient of the abdominal signal was taken
the cases are the combinations of abdominal
as the original input of the adaptive filtering system,
signals and thoracic signals. The algorithm was
and the coefficient of the thoracic signal was taken as
implemented in Matlab (R2009a).
the reference input.
The simulated data were generated by a
(3) Correlations of the processed wavelet
software tool called UG DigiScope [37]. Nine ECG
coefficients were computed. The threshold was set
signals were generated with the heart rate of 64, 76,
and noise components were removed with the SSNF
80,
32177 84, 94, 120, 130, 140 and 154 beats per
algorithm.
minute. The sampling frequency was 250 Hz, and
¼ M
the acquisition time was 30 s. The ECG signals with were included: 5 abdominal signals and 3 thoracic
the heart rate from 64 to 94 were used as maternal signals. The sampling rate was 250 Hz, and the
ECG, and the others as fetal ECG. The simulated acquisition time was 10 s. In fetal ECG extraction,
fetal ECG amplitude is half of the maternal ECG. except the 4th channel which is unstable and
Simulated abdominal signals were obtained by cannot be used for algorithm verification tests, the
addition of simulated maternal ECG and fetal ECG. 1st to the 3rd channels and the
The simulated abdominal signals were used as the
5th channel of abdominal signals were used
original input, while the maternal ECGs were used as
as the original input, while the 6th to the 8th
the reference input.
channels of thoracic signals were used as the
The clinical data were from the Database for reference input.
the Identification of Systems (DaISy) by Lathauwer
Results and analysis
[38]. This well-known database contains ECGs
measured from a pregnant woman. The first 1000 The simulating experimental results with the
points were used to plot the ECG waveform which heart rate of 80 and 140 beats per minute are
is shown in Fig. 3. Eight channels of signals shown in Fig. 2. The three wave-
All the n elements in D(n) were compared simulated maternal ECG, and the fetal ECG
with the threshold Thr. If any element ni (1 r ni r extracted. It can be observed that the proposed
n) satisfies D(ni) 4 Thr or D(ni) o Thr, the method could remove maternal ECG and extract fetal
corresponding element x(ni) will be detected as an ECG efficiently. However, it should be noted that
R-peak. the first waveform of fetal ECG is not complete. The
reason is that the adaptive filter has to adjust the
Quantitative performance can be
coefficients before the stable state. Other simulated
evaluated by the signal to noise ratio (SNR) based
data with different heart rates were used to extract
on eigenvalue analysis and cross correlation
fetal ECG, and all of them can extract fetal ECG
from the simulated abdominal signals.
32178
Channels (ch) 1–5 are abdominal signals while
Channels 6–8 are thoracic signals.
In the clinical experiments, the LMS

algorithm was directly applied to the same data to
compare the performance. Results of both the LMS
algorithm and the proposed algorithm are shown in
Fig. 4. Fig. 4(a) shows the extraction results of
LMS while Fig. 4 (b) shows the results of our
method. The three waveforms from top to bottom
are the abdominal signal, the thoracic signal, and the
Fig. 2. Experimental results of simulated fetal ECG extracted. The circle on the third
data. Three waveforms from top to bottom are waveform shows the apex of each R wave. It can be
simulated abdominal signal, simulated maternal observed in Fig. 4 that the result of fetal ECG
thoracic electrocardiogram (ECG), and the extraction is markedly improved, and both maternal
extracted fetal ECG (FECG). ECG and other disturbance are eliminated. For the
LMS algorithm, maternal components are still
visible in the output fetal ECG as indicated by
ellipses. Besides, the second QRS-wave of fetal ECG
in Fig. 4(a) shows improvement over Fig. 4(b),
in case of superposition R-peaks of fetal ECG
and maternal ECG.
Further verification for the algorithm

performance is conducted by calculating SNRs
based on eigenvalues (SNR-E) and cross correlation
coefficients (SNR-C). The comparison between the
LMS algorithm and our algorithm with respect to
SNR-E and SNR-C are shown in Figs. 5 and 6,
respectively, both using the 12 cases of clinical data.
It can be seen in Figs. 5 and 6 that for Cases 1–9,
SNR- E and SNR-C of the proposed algorithm are
Fig. 3. Eight channels of electrocardiogram obviously higher than those of the LMS algorithm.
(ECG) signals collected from a pregnant woman.
32179 Among them Case 8 has the best
Fig. 5. Comparison of the signal to noise

ratio (SNR) based on eigenvalues (SNR-E)
of the least mean square (LMS) algorithm

and our algorithm.
performance, which is a combination of

Channel 3 abdominal signal with Channel 7
thoracic signal. For Cases 10–12, the differences
Fig. 4. Comparison of experimental results
of SNR-E and SNR-C of the two algorithms are
of clinical data (a) the result of our algorithm and
small. These three cases are combinations of
(b) the result of the least mean square (LMS)
Channel 5 abdominal signal with Channels 6–8
algorithm. Three waveforms in each figure from top
thoracic signals. Because Channel 5 abdominal
to bottom are Channel (ch) 1 abdominal signal,
signal has a low component of fetal ECG, the
Channel 8 thoracic signal, and the extracted fetal
extraction results of the three cases are poorer than
electrocardiogram (FECG). Circles in the FECG
the other cases. Nevertheless, for Cases 10–12,
waveform show the apex of R waves while
SNR-E and SNR-C of our algorithms are still
ellipses show the disadvantage of the LMS result.
slightly larger than that of the LMS algorithm. To
sum up, these results show that our new algorithm
improves performance considerably over the LMS
algorithm.
32180
reference input is the thoracic signal. Twenty cases

of simulated data and 12 cases of clinical data were
used. Experimental results showed that the proposed
method outperforms the LMS adaptive filtering
algorithm: (1) it shows improvement in case of
superposition R-peaks of fetal ECG and maternal
ECG; (2) noise disturbance is eliminated by
incorporating the SSNF algorithm and the extracted
waveform is more stable; and (3) the
performance has been proven quantitatively by
Fig. 6. Comparison of the signal to noise
SNR calculation. The results indicated that the
ratio (SNR) based on cross correlation coefficients
proposed algorithm can be used for extracting
(SNR-C) of the least mean square (LMS) algorithm
fetal ECG from abdominal signals. Future work can
and our algorithm.
be conducted on real-time implementation of the
algorithm and further clinical validation.
One limitation of this work is that it has not The authors would like to thank the
been validated on pregnant women in the clinic. To anonymous reviewers for their insightful comments
verify its clinical feasibility, the proposed method and recommendations.
must be tested clinically and improved corre-

References
spondingly, and it seems that comparing the method
with available commercial ECG systems is
desired. Another limitation is that the current
[1] W. Jia, C. Yang, G. Zhong, M. Zhou,
processing is off-line. These issues must be resolved
S. Wu, Fetal ECG extraction based on
to promote the method toward clinical application.
adaptive linear neural network, in: 3rd
International Conference on Biomedi- cal
Conclusions
Engineering and Informatics, Yantai, China,
This paper presented a new method for fetal 16–18 October 2010, pp. 889–
ECG extraction combining wavelet decomposition,
902.
the LMS adaptive filtering algorithm, and the
[2] L.D. Lathauwer, B.D. Moor, J.
SSNF algorithm. The algorithm has two inputs. The
original input is Vandewalle,
the abdominal signal, while the 32181 Fetal electrocardiogram
extraction by blind source subspace biomedical signal extraction, Sci. China Ser.
separation, IEEE Trans. Biomed. Eng. 47 F 52 (2009) 1863–1874.
(2000) [8] A. Al-Zaben, A. Al-Smadi, Extraction of
567–572. foetal ECG by combination of singular value
[3] K. Assaleh, Adaptive neuro-fuzzy decomposition and neuro-fuzzy inference
inference systems for extracting fetal system, Phys. Med. Biol. 51 (2006) 137–143.
electrocardiogram, in: International [9] P.P. Kanjilal, S. Palit, G. Saha, Fetal
Symposium on Signal Processing and ECG extraction from single-channel
Information Technology, Vancouver, BC, maternal ECG using singular value
August 2006, pp. 122–126. decomposition, IEEE Trans. Biomed. Eng.
[4] S. Puthusserypady, Extraction of fetal 44 (1997) 51–59.
electrocardiogram using H(infinity) [10] Y. Senim, A. Atasoy, Performance
adaptive algorithms, Med. Biol. Eng. evaluation of nonparametric ICA algorithm
Comput. 45 (2007) 927–937. for fetal ECG extraction, Turk. J. Electr. Eng.
[5] M. Ungureanu, J.W. Bergmans, S.G. Comput. Sci. 19 (2011) 657–666.
Oei, R. Strungaru, Fetal ECG extraction [11] J.L. Camargo-Olivares, R. Martin-
during labor using an adaptive maternal beat Clemente, S. Hornillo-Mellado, M.M. Elena,
subtraction technique, Biomed. Tech. (Berl.) I. Román, The maternal abdominal ECG as
52 (2007) 56–60. input to MICA in the fetal ECG extraction
[6] H. Zhang, Z. Shi, C. Guo, E. Feng, Semi- problem, IEEE Signal Process. Lett. 18 (2011)
blind source extraction algorithm for 161–164.
fetal electrocardiogram based on [12] F.S. Najafabadi, E. Zahedi, M.A. Mohd
generalized autocorrelations and reference Ali, Fetal heart rate monitoring based on
signals, J. Comput. Appl. Math. 223 (2009) independent component analysis, Comput.
409–420. Biol. Med. 36 (2006) 241–252.
[7] Y. Ye, P.C.Y. Sheu, J. Zeng, G. Wang, [13] M. Hasan, M. Reaz, Hardware
K. Lu, An efficient semi-blind source prototyping of neural network based fetal
extraction algorithm and its applications to 32182
electrocardiogram extraction, Meas. Sci. [19] D. Tarălungă, W. Wolf, R. Strungaru,

Rev. 12 (2012) 52–55. M. Ungureanu, Abdominal signal proces-
[14] X.J. Pu, X.P. Zeng, Y.J. Chen, W. Yu, L. sing: fetal ECG extraction by combining
Han, J. Cheng, Fetal electrocardiogram ESC and ICA methods, in: 4th European
extraction based on radial basis function Conference of the International Federation
neural networks, J. Chongqing Univ. for Medical and Biolo- gical Engineering,
32 (2009) 111–115. Antwerp, Belgium, 23–27 November 2008,
[15] Z. Wei, L. Hongxing, C. Jianchun, pp. 1196–1199.
Adaptive filtering in phase space for foetal [20] Y. Zeng, S. Liu, J. Zhong, Extraction of
electrocardiogram estimation from an fetal ECG signal via adaptive noise
abdominal electrocardiogram signal and a cancellation approach, in: 2nd International
thoracic electrocardiogram signal, IET Signal Conference on Bioinformatics and
Process. 6 (2010) 171–177. Biomedical Engineering, Shanghai, China,
[16] T.Q.D. Khoa, H.H.M. Tam, V.V. Toi, 16–18 May 2008, pp. 2270–2273. [21] K.B.
Extracting fetal electrocardiogram from Kim, D.K. Hsu, D.J. Barnard, Estimation of
being pregnancy based on nonlinear porosity content of composite
projection, Math. Probl. Eng. (2012) 13. materials by applying discrete wavelet
(Article ID 439084). transform to ultrasonic backscattered signal,
[17] R. Martin-Clemente, J.L. Camargo- NDT E Int. 56 (2013) 10–16.
Olivares, S. Hornillo-Mellado, M.M. Elena, [22] P. Kumar, D. Agnihotri, Biosignal
I. Román, Fast technique for noninvasive denoising via wavelet thresholds, IETE J.
fetal ECG extraction, IEEE Trans. Biomed. Res.
Eng. 58 (2011) 227–230. 56 (2010) 132–138.
[18] M.G. Jafari, J.A. Chambers, Fetal [23] G.P. Nason, B.W. Silverman, The
electrocardiogram extraction by sequential stationary wavelet transform and some
source separation in the wavelet domain, statistical applications in wavelet and
IEEE Trans. Biomed. Eng. 52 (2005) statistics, Lect. Notes Stat. 103 (1995)
32183
390–400. 281–299.
[24] A. Khamene, S. Negahdaripour, A new [29] Y. Yue, The application of a new

method for the extraction of fetal ECG threshold function in wavelet denoising,
from the composite abdominal signal, IEEE (M.E.
Trans. Biomed. Eng. 47 (2000) Thesis), Changchun University of
507–516. Technology, China, April 2011.
[25] M.G. Jafari, J.A. Chambers, Fetal [30] S. Mallat, W.L. Hwang, Singularity
electrocardiogram extraction by sequential detection and processing with wavelets, IEEE
source separation in the wavelet domain, Trans. Inf. Theory 38 (1992) 617–643.
IEEE Trans. Biomed. Eng. 52 (2005) [31] Y. Xu, J.B. Weaver, D.M. Healy, J.
390–400. Lu, Wavelet transform domain filters:
[26] W. Jia, S. Wu, Y. Bai, Fetal ECG a spatially selective noise filtration
extraction based on wavelet transform technique, IEEE Trans. Image Process.
modulus maximum algorithm, in: 5th 3 (1994) 747–758.
International Conference on Cooperation [32] D.L. Donoho, De-noising by soft-
and Promotion of Information Resources in threshoding, IEEE Trans. Inf. Theory 41
Science and Technology, Beijing, China, (1995)
27–29 November 2010, pp. 528–532. 617–627.
[27] C. Guo, M. Li, An improved image [33] D. Jin, Y. Wang, W. Wang, A
denoising algorithm based on wavelet wavelet space based approach for Doppler
transform modulus maximum, in: ultrasound blood signals separation, Chin. J.
International Conference on Computer Acoust. 26 (2007) 261–264.
Application and System Modeling,
Taiyuan, China, 22–24 October 2010, pp.
296–299.
[28] S. Mallat, Characterization of signals
from multiscale edges, IEEE Trans. Pattern
Anal. Mach. Intell. 14 (1992) 710–732.
32184
32185
32186

M.Vithya, S.Meenakshi: Noise Cancellation in Fetal ECG Extraction Using SSNF Algorithm

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

M.Vithya, S.Meenakshi: Noise Cancellation in Fetal ECG Extraction Using SSNF Algorithm

Uploaded by

Copyright:

Available Formats

International Journal of Applied Engineering Research, ISSN 0973-4562 Vol. 10 No.

Noise Cancellation in Fetal ECG Extraction Using SSNF Algorithm

Extracting clean fetal electrocardiogram (ECG) signals is very important in fetal

Fetal electrocardiogram (ECG) waveform A number of techniques have been proposed

Adaptive ﬁltering algorithm When wavelet decomposition is performed

Fig. 1. Adaptive ﬁltering conﬁguration.

Wavelet coefﬁcients related to signals can be extracted by M 1 T

(4) The processed wavelet coefﬁcients

In the clinical experiments, the LMS

Further veriﬁcation for the algorithm

Fig. 5. Comparison of the signal to noise

of the least mean square (LMS) algorithm

performance, which is a combination of

reference input is the thoracic signal. Twenty cases

verify its clinical feasibility, the proposed method and recommendations.

must be tested clinically and improved corre-

electrocardiogram extraction, Meas. Sci. [19] D. Tarălungă, W. Wolf, R. Strungaru,

[24] A. Khamene, S. Negahdaripour, A new [29] Y. Yue, The application of a new

You might also like