EEG signal recognition for brain word interface using wavelet decomposition

Abstract

A simple brain word dictionary (BWD) system using wavelet decomposition to form feature sets is developed. A BWD is an essential tool in the rehabilitation of paralyzed individuals which converts the brain EEG signals into audio words. A feed forward neural network classifier is proposed to classify ten simple words. EEG signals acquired from two subjects are used in the experiments. Performance of the single trial analysis has an average recognition rate of 87.7%.