Extraction of Electrocardiography Artefacts From Surface Electromyography Signals Using Continuous Wavelet Transform
The study of biosignals such as Electromyography (EMG) and Electrocardiography (ECG) signals is significant within the physiological and medical field for diagnosis and rehabilitation purposes. The extraction of features from these signals by analysing them can be used to understand the possibilities of health and physiological status further. The presence of ECG contamination within surface Electromyography signal (sEMG) has been a problem when analysing the muscle activity. It has been shown that ECG contamination of the EMG signal in muscles of the trunk can influence time-based (amplitude) and frequency-based measures of the EMG signal. This has important implications when interpreting the amount of muscle recruitment and the level of fatigue of trunk musculature. To date, there is little information showing the degree of contamination of the ECG when the trunk musculature is contracting at different intensities. There is no clear information regarding whether the ECG signal can be effectively extracted from the EMG to detect other physiological variables such as heart rate and heart rate variability when the trunk muscles are contracting at different intensities. The significance of the presence of ECG artefacts within sEMG signals at various levels of muscular activity is seen to be diminishing as the intensity of EMG signals increases as the activity across the muscle increases. The different levels of EMG activity are measured by monitoring the degree of Maximum Voluntary Contractions (MVIC) of the muscle. This research involved the investigation of the ECG artefacts within sEMG signals that were obtained from the lower and upper lumbar erector spinae (ES) muscles of the back. This research involved the extraction of the ECG artefacts from sEMG signals using Continuous Wavelet Transforms (CWT) with thresholding. This technique was applied to sEMG signals obtained from the lower lumbar erector spinae muscles during different levels of static contractions, spanning from 5%-50% of Maximum Voluntary Isometric Contraction (MVIC) at 5% intervals. Surface EMG signals collected across a group of healthy participants within the age bracket of 18-35 with no previous history of back injury or surgeries within the last year. This research explored the difference in signal properties before and after extracting the ECG signals from the sEMG signals using the CWT. The CWT provides a scalogram plot that indicates the power intensity at each scale for the extracted signal within the time domain. The CWT scalogram can be replotted to show the corresponding pseudo-frequency-time based spectrum plot. These plots were used to provide the scale values most suitable for the ECG extraction from the sEMG signals. Using the selected CWT scales, thresholding of these wavelet coefficients was performed prior to the reconstruction of the extracted information to diminish the presence of EMG signals within the reconstructed ECG signals. This technique was used due to the similarity in nature of EMG signals to that of white Gaussian noise. The reconstructed ECG signal is cross verified with an independent 3-lead ECG recording that was collected simultaneously from the same participant during the EMG data collection. The significance of the ECG signal before and after extracting the ECG signal from the sEMG signal was validated using Fourier power spectrums and finding the median frequencies within the selected time segment. The ECG extraction from sEMG signals using the CWT technique with thresholding was shown to have been successful at lower percentages (5%-20% of MVIC) and was able to have extracted ECG components significantly lowering the median frequency of the EMG signal after the removal of ECG signal. The research showed results demonstrating the extraction of ECG signals from sEMG signals collected from the back muscles at low values of MVIC.