A Brain-Computer Interface (BCI) System Based on Auditory Stream Segregation

Abstract

An auditory brain-computer interface (BCI) which detects event-related potentials (ERPs) elicited by selective attention to one of the tone streams was proposed. Two frequency oddball tone sequences with different tone frequency ranges were alternately presented to subjects, and were perceived by subjects as two kinds of segregated streams. Event-related potentials elicited by two kinds of deviant tones were classified by linear discriminant analysis (LDA) to find the stream subjects paid selective attention to. Experiments with six subjects have shown that this system could realize binary selection from two tone streams.