A novel method for automatic ICi selection in EEG signals

Abstract

Advancements in EEG signal processing yields tremendous research in the field of BCI. Online based BCI fails in selecting independent components from different brain regions. Unknown order of the independent components and the random weight matrix used in repeated ICA trainings may leads to a different ICA result. This paper highlights a new idea of automatic ICi selection by taking an average of particular brain regions which resolves the problem of online BCI. The proposed method has been tested in EEG datasets such as .SET, .SMA which succeeds in selecting reference ICi.