Channel selection based on amplitude and phase characteristics for P300-based brain-computer interface

Abstract

Brain-Computer Interface (BCI) paradigms communicate with external world through specific electrode positions. As the oscillatory brain activity varies with different individuals, the prior experience is not enough to select the optimal channels for pattern recognition. Most channel selections are based on amplitude features. In this paper, an approach for channel selection that combines amplitude and phase features is proposed. Applied to the data set IIb of BCI Competition 2003, the method for channel selection achieved an satisfying accuracy that only one letter was predicted as wrong among thirty-one letters.