Towards Single Trial Decoding Of Cortical-Muscular Activities

This paper exploited the single trial decoding of cortical-muscular activities (CMAs). The CMAs were measured by acquiring Electromyographic (EMG) signal, and Electroencephalographic (EEG) signal. We focused on CMAs related to sustained muscular contractions (SMC), and investigated on the classification performance of four types of CMAs with different combinations of three types of features. The four types of CMAs were relaxing, conducting precise hand operation without distractive mental task, conducting precise hand operation with distractive mental task, and conducting rough hand operation. The three types of features were combined root mean square (RMS) and wavelet packet decomposition (WPD) feature of EMG signals, common pattern spatial (CSP) filtered WPD feature of EEG signals, and cortical-muscular coherence (CMC) feature of EMG-EEG signals, which were noted as EMG features, EEG features, and CMC features respectively. In total of 5 subjects participated in the experiments. According to experimenttal results, we concluded that the classification performance of myoelectric controller can be significantly improved (p<0.05, ANOVA) by the combination of EEG and EMG features and the combination of CMC and EMG features, compared with sole EMG features. The improvements were 9.83% and 6.06% respectively.