Dynamic Analysis of Motor Imagery EEG Using Kurtosis Based Independent Component Analysis

This paper investigates the possibility of using independent component analysis (ICA) to analyze EEG dynamics of motor imagery. Considering the non-stationary characteristics of the motor imagery EEG, the online Independent Component Analysis by kurtosis maximization is proposed to detect the mu rhythm changes during left and right hand movement imagination. The experiment results show that the kurtosis based online ICA can concentrate the mu rhythm in raw EEG channels to one output channel, but the batch ICA fails to do that. The study in this paper also show that the elements of dynamic mixing matrix are more sensitive to mu rhythm dynamic changes, which means the parameters of dynamic mixing model online ICA can be used to monitor the dynamic changes of mu rhythm during motor imaginations. The results in this paper demonstrate that the online ICA may be a promising tool for the analysis of EEG dynamics.