Spectral analysis of lower limb EMG activity in individuals with motor complete SCI during standing with epidural stimulation

The EMG signals have been recorded from lower limb muscles during standing in individuals with complete spinal cord injury implanted with spinal cord epidural stimulation. In this study, the EMG signals are classified based on the quality of standing using spectraldomain features obtained by three frequency transformation methods: Fast Fourier Transform (FFT), Short-time Fourier Transform (STFT) and Continuous Wavelet Transform (CWT). The k-nearest neighbor method was used for the classification task. It is shown that CWT features have the highest classification accuracy (~ 90%) compared to STFT and FFT. This suggests that CWT provides more in-depth assessment of time-frequency content of the EMG signals that can discriminate between assisted and independent standing performances.