Word Recognition using speech-imagery EEG

Abstract

Previous research has suggested that humans manipulate the machine using their electroencephalogram called BCI (Brain Computer Interface). However, there are not existed effective methods for speech-imagery recognition. In this report, we propose the word recognition method using kurtosis and power spectral entropy extracted from EEG signal of continuously speech-imagery. The word recognition of ten numbers was conducted in our experiment, the average accuracy of 82.6% was obtained. It is clear that these features are effective for word recognition using EEG signal of continuously speech-imagery.