Background: Electrocardiogram (ECG) signals provide confirmation about the cardiac
rhythmic deviations and its function. But throughout the acquisition of Electrocardiogram signal from
the human body, it becomes distorted with artifacts like 50 Hz power line interference, wandering of
base line, respiratory signals and muscle artifacts. In spite of the noises in ECG, 50 Hz power line
interference will relentlessly disturb the nature of the ECG signal. In 1990, Huang presented the algorithm
called Empirical Mode Decomposition (EMD). It is an adaptive recursive decomposing algorithm
that disintegrates the signal into different levels called Intrinsic Mode Functions (IMF). EMD has
the limitations like sensitivity to noise and sampling. This problem is only partially addressed by other
techniques like synchrosqueezing and empirical wavelets. Variational mode decomposition (VMD) is
a non-recursive method in which the modes are extracted concomitantly. This technique is more robust
to sampling and noise.
Methods: In this paper novel VMD based subtraction techniques are developed and its denoising
performances are analysed. This proposed that method outperforms other denoising methods like
notch filtering and wavelet transform. Thus the characteristic features of ECG signal are undisturbed
by preserving its originality.
Results: The comparative performance of various methods in terms of the performance parameters
like MSE, SNR, PRD and PSNR is carried out. The SNR level is stable in all noise levels. The SNR
is varying from 19dB to25dB using hard and soft thresholding techniques. But it is almost double
when our proposed approach is applied. The SNR value of direct method is ranging from 38dB to
43dB with a minimum MSE of 0.004 at 5% noise level. The noise band of 48 Hz to 51Hz is
separated from the VMF level so that the desired ECG signal component is undestroyed. Thus
compared to direct subtraction indirect subtraction technique performs well. The SNR value ranging
from a range of 54dBto 55dB is obtained using an indirect method. Highest SNR value of 55.98 dB
is obtained at 5% noise level and a very low MSE value is also obtained.
Conclusion: This paper provides a novel method in the area of denoising Biosignals. Two
techniques are proposed for denoising such as direct subtraction and indirect subtraction techniques.
A comparative study is performed to study the performance of the proposed method. The
parameters used for comparison are MSE, SNR, PRD and PSNR. Indirect subtraction technique
outperforms the direct subtraction method. An improved SNR value and minimized MSE indicates
the performance of denoising method. Thus VMD based subtraction techniques give better denoised
signals even though the decomposition process takes time.