Decomposition of EMG signals using time-frequency features

The decomposition of intramuscular myoelectric (EMG) signals can be considered as a classification problem. The main effects which decrease the classification performance are Motor Unit Action Potential (MUAP) shimmer and overlapping MUAPs. In this paper we show how time-frequency information can be extracted to reduce MUAP shimmer and propose a criterion to detect overlapping MUAPs. Because of the information extraction and detection of compound MUAPs, the classification problem can be reduced to a detection problem of highly isolated cluster points. Tests with EMG recordings yield very good results.