An Auditory Brain-Computer Interface Using Active Mental Response

This study proposes a novel auditory brain-computer interface paradigm, which allows the subject to mentally select a target among a random sequence of spoken digits. The subject's voluntary recognition of the property of the target digits enhances the discriminability between brain responses to target and nontarget digits. EEG data from 14 subjects has shown that the amplitude of N2 and the late positive component (LPC) elicited by target digits was significantly higher than that of nontarget ones. Three classification methods, i.e., N2/LPC area comparison, Fisher discriminant analysis and support vector machine (SVM), were adopted to assess the target detection accuracy using EEG data from a single electrode. For SVM classification, a mean accuracy of 85% was achieved with five trials averaged. This new paradigm could be useful for locked-in patients with vision impairment.