An Improved EEG Feature Extraction Method Based on Quantum Particle Swarm Optimizer Algorithm

Abstract

Feature extraction is a very important step in P300based brain-computer interface(BCI). Aiming at the large amount ofelectroencephalogram(EEG) data used in BCI, a method which combined quantum particleswarm optimizer(QPSO) algorithm with independent component analysis(ICA) technique was putforward for P300 extraction.It initialized several particles in the feasible domain of de-mixing matrixw, making them parallel search, andfinally reached the global convergence by the information communication betweenthe current particle and the global optimum particle. The method also optimizedthe search strategy using quantum computing to impelICA iterationto converge faster. It was test on real EEG datasetand a simple linear classifier was employed.The recognition accuracy arrivedat 94.5% with 15 times averaged. The results showed thatthe proposed method could extract P300 rapidly and efficiently.It provided a better way to research and develop BCI systemfurther.