A template subtraction method for reducing electrocardiographic artifacts in EMG signals of low intensity

The electromyogram (EMG) has been used in several studies for different areas. Whenever recorded on trunk muscles and close to heart, the EMG may be contaminated by electrocardiographic signal (ECG), which may hamper information based on EMG measures, by either decreasing the median frequency (MDF) or increasing the root mean square value (RMS). Therefore, ECG removal from this contaminated EMG is an issue, but the signal processing task is challenging due to the spectral overlapping of signals. The Butterworth high-pass filter with a 30 Hz cutoff frequency has been considered a suitable removal technique in the literature. However, the frequency band below 30 Hz is strongly attenuated, leading thus to information loss on the EMG. In order to mitigate this problem, a simple template subtraction method is proposed, in which the template is based on the contaminated signal itself. Real contamination and emulated mixtures based in real signals were used and the MDF, RMS, correlation coefficient and Kulback-Leibler divergence were analyzed as performance measures. The performance of the proposed method was compared to a Butterworth high-pass filter and it is shown that the template subtraction method has preserved better the EMG information.