Recognition of arm-movement electroencephalography (EEG) using motor-related intrinsic mode functions filtering and cross mutual information: World Congress on Medical Physics and Biomedical Engineering: Neuroengineering, Neural Systems, Rehabilitation and Prosthetics

In this paper, we propose trial-specific and subject-specific filters to extract the motor-related compartment for the recognition of EEG signals induced by the left- or right-arm movement. Such motor-related filters are the intrinsic mode functions (IMF), which were produced by the decomposition of signals in C3 or C4 motor channels using the empirical mode decomposition (EMD), with the peak frequency pertaining to the mu rhythm within alpha band (8-12Hz) or beta band (16-25Hz). After these trial-specific and subject-specific filters were applied on all channels, the cross mutual information (CMI) of filtered signals between any two channels was computed. The average classification rates for five healthy subjects obtained from the proposed filters related to the alpha and beta bandpass filtering with whole-brain CMI maps were 77.4% and 88.3%, respectively, which were superior to the 72.2%, and 77.7% obtained from the same filtering but with only CMI vectors related to motor signals in C3 and C4, respectively.