Denoising of ECG and Estimation of the Baseline Using Group Sparse Regularization and Some Advanced Filtering Techniques

Abstract

In this context, the sparse features of the particular signal are taken into consideration to exhibit the technique of sparse optimization. This type of approach is derived from the traditional approach known as total variation. This traditional approach is the old one used to reduce the noise in a particular signal in order to rescue the sharp edges of the signal. In this context, majorization and minimization are the major keys to getting the desired noise-free signal. Along with that, most advanced filters like median filters and moving average filters are considered in order to gain a noise-free signal, and the signal will be more vulnerable to identification. Using the filters discussed above, the signal can be determined in the high SNR ranges as well.

Introduction

I. INTRODUCTION

In order to know the exact condition of the heart and the illness associated with it, the electrocardiogram signal is the only criteria. Due to technological advancements in the medical field, remote ECG monitoring is now possible everywhere, and this type of improvement in this particular field helps in the automatic detection of heart disorders. Signals from the ECG are polluted with unwanted disturbances and noises such as baseline wander, artefacts, and other noises from outsource. These are all together responsible for causing the disturbance in the signal. So, eliminating these polluted signals from the pure signals is very much needed in order to know the exact condition of the heart and to discover the illness incorporated with the heart. The ECG signal consists of different components, like the P-wave, QRS complex, and T-wave. The P-wave is used for starting the atrial contraction; it is the depolarization of the atria. This is followed by the contraction of the ventricles, or depolarization, and then the T-wave is used for the repolarization of the ventricles.

A moving average filter, also alternatively named rolling average or running average, involves averaging various distinct subsets of a complete data set. It is also sometimes spelled moving mean or rolling mean. The subset of the number series that is initially fixed is averaged in order to produce the moving average first component. Then the subset, which is fixed initially, is altered by "shifting forward," and it includes the next value in the subset.

II. LITERATURE REVIEW

1. M. Z. U. Rahman, R. A. Shaik and D. V. R. K. Reddy, "Efficient and Simplified Adaptive Noise Cancelers for ECG Sensor Based Remote Health Monitoring," in IEEE Sensors Journal, vol. 12, no. 3, pp. 566-573, March 2012.

The contaminated noise in electrocardiogram data uses adaptive filters that are based on error nonlinearity, which are more complex. The suggested one will be more suitable for more advanced applications like biometry, which has high signal-to-noise ratios. These types of approaches perform least mean square (LMS) realisations by using addition and shift operations. As for the comparisons, the suggested one will perform better for all the parameters mentioned above.

Summary: Noise in ECG signals further improved.

2. Kabir, Md. Ashfanoor and Celia Shahnaz. “Denoising of ECG signals based on noise reduction algorithms in EMD and wavelet domains.” Biomed. Signal Process. Control. 7 (2012): 481-489.

In this context, the EMD-based denoising of the ECG is introduced, which differs from the conventional approaches. In this context, they have used the EMD domain to eliminate the noise from IMFs, which helps retain the QRS complex signal. This method will not pay attention to the number of IMFs containing QRS complexes or the noise associated with them. The particular signal here is changed into the DWT domain, which helps to construct the clear one of the particular ECG signal. They have compared four MIT-BIH arrhythmia database simulations; thus, from the resulting output, the noise is further decreased, and there is high scope for getting a more accurate denoised signal.

Summary: New ECG denoising approach designed.

3. L. Smital, M. Vítek, J. Kozumplík and I. Provazník, "Adaptive Wavelet Wiener Filtering of ECG Signals," in IEEE Transactions on Biomedical Engineering, vol. 60, no. 2, pp. 437-445, Feb. 2013.

Signals that are extracted from the standard CSE database and noised manually are captured for testing at 500 Hz. They incorporated the signal with white Gaussian noise and altered its power spectrum. The reason for the incorporation is to match the power spectrum model of the standard EMG signal to create the illusion of fake interference. Choosing the accurate filter banks and also choosing parameters of the Wiener filter based on the signal-to-noise ratio are their main aims. They have used some filtering settings that adjust the filtering parameters with the amount of interference in the input signal in order to increase the performance of the filtering.

Summary: Reduction of broadband myopotentials (EMG) in ECG signals further improved.

4. S. Poungponsri and X. -H. Yu, "Electrocardiogram (ECG) signal modeling and noise reduction using wavelet neural networks," 2009 IEEE International Conference on Automation and Logistics, 2009, pp. 394-398.

In this context, in order to eliminate the noise, the wavelet neural network approach is introduced. This is the hybrid approach, which is a combination of adaptive diversity learning particle swarm optimisation (ADLPSO) and the gradient descent optimisation method.

From the resultant simulations, the approach they describe here is worth using and also efficiently eliminates the high-frequency noise.

Summary: WNN is implemented.

5. Xiong, Peng, Hongrui Wang, Ming Liu, Feng Lin, Zengguang Hou and Xiuling Liu. “A stacked contractive denoising auto-encoder for ECG signal denoising.” Physiological Measurement 37 (2016): 2214 - 2230.

In this paper, in order to eliminate the noise from the particular signal, a new approach called contractive denoising auto encoder (CDAE) is described, which creates a deep neural network for the reduction of the noise. It increases the ECG signal expressions by using multilevel feature extraction, which is based on the Frobenius norm of the Jacobean matrix. The MIT-BIH database is used for the approach. The novel method proposed here will lead to improvements in the signal-to-noise ratio and root mean square error.

Summary: Denoising technique is further improved

Conclusion

The unwanted noise in the signal is removed by the group sparse optimization method and also by concatenating the simplified filters like the median and the moving average filter. The proposed method is applicable at high signal-to-noise ratio (SNR) values. The minimization and majorization algorithm is the main lead in this type of approach, and the BW and other noises are successfully eliminated or reduced, and we successfully obtained a clear ECG signal. By concatenating with other filters, the denoising will be more effective, and a more accurate signal can also be obtained.

References

[1] M. Z. U. Rahman, R. A. Shaik and D. V. R. K. Reddy, \"Efficient and Simplified Adaptive Noise Cancelers for ECG Sensor Based Remote Health Monitoring,\" in IEEE Sensors Journal, vol. 12, no. 3, pp. 566-573, March 2012. [2] Kabir, Md. Ashfanoor and Celia Shahnaz. “Denoising of ECG signals based on noise reduction algorithms in EMD and wavelet domains.” Biomed. Signal Process. Control. 7 (2012): 481-489. [3] L. Smital, M. Vítek, J. Kozumplík and I. Provazník, \"Adaptive Wavelet Wiener Filtering of ECG Signals,\" in IEEE Transactions on Biomedical Engineering, vol. 60, no. 2, pp. 437-445, Feb. 2013. [4] S. Poungponsri and X. -H. Yu, \"Electrocardiogram (ECG) signal modeling and noise reduction using wavelet neural networks,\" 2009 IEEE International Conference on Automation and Logistics, 2009, pp. 394-398 [5] Xiong, Peng, Hongrui Wang, Ming Liu, Feng Lin, Zengguang Hou and Xiuling Liu. “A stacked contractive denoising auto-encoder for ECG signal denoising.” Physiological Measurement 37 (2016): 2214 - 2230. [6] ECG Baseline Estimation and Denoising With Group Sparse Regularization. HAO SHI , RUIXIA LIU , CHANGFANG CHEN , MINGLEI SHU , AND YINGLONG WANG

Copyright

Copyright © 2023 A. Rajani, Ch. Ramya Rekha. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.