Evoked EEG Feature Extraction Technique Based on Independent Component Clustering Analysis

A new method based on independent component clustering analysis(ICCA) for evoked EEG(electroencephalograph) feature extraction is developed,which uses the higher inter trail similarity among evoked components and combines Infomax with K-means algorithm to separate and extract evoked components.This method overcomes the difficulty of evoked component recognition in traditional independent component decomposition methods and shows the advantage in automatic processing on high dimensional(EEG) data evoked by multiple stimuli.Using ICCA analysis on EEG signals during imaginary upper extremity movement,the result shows that ICCA can automatically separate background noise and extract evoked components,thus increases the Fisher separability dramatically and achieves a better recognition performance.This study demonstrates that ICCA is fit for cognitive behavioural EEG signal analysis and is worthy of further research.