Combination of the Channel Dimension Expansion and FastICA for the Recognition of BCI Motor Imagery EEG

As a fast algorithm of independent analysis (ICA), FastICA is drawing attention due to its rapid convergence speed. Based on the application of FastICA algorithm with the criterion of neg-entropy to the recognition of motor imagery electroencephalogram (EEG), and the characteristics of ICA, this paper designs the experimental process for data treatment. Concerning the problem of few channels in the data set III in the BCI Competition 2003, a channel dimension expansion algorithm is put forward in this paper, which can exponentially increase the number of similar channels and provide more brain electricityinformation without increasing the number of electrodeplates to be collected. In this paper, the channel dimension expansion is combined with the algorithm of FastICA to be applied to the data treatment of BCI Competition. The experiment results suggest that the channel dimension expansion algorithm can improve the classification accuracy of FastICA, and that the fast information treatment of FastICAcan make up for the time-consuming side of the channel dimension expansion algorithm.