Surface electromyography denoising using empirical mode decomposition during gait

Recently empirical mode decomposition (EMD) has been proposed to denoise the surface electromyography (EMG) signals. The EMD decomposes the signal in intrinsic mode functions (IMFs). The selection of IMFs that contain noise is usually selected based on noise thresholds. However, threshold amplitude is arbitrary. To solve the problem of the arbitrary threshold, we propose a method-based IMFs reconstruction pattern with higher kurtosis. The power between 48-51 Hz (i.e. electromagnetic noise), the duration of burst activation and the amplitude of signal were compared between EMD and a traditional filter (i.e. Butterworth filter 4th order, bandpass 20-400 Hz) during gait in a treadmill. Our results show that EMD method reduces the power of electromagnetic noise and the duration of burst activation, without a change in EMG amplitude. The selection pattern of IMFs with higher kurtosis could be a good alternative to the traditional method of denoise EMG.