A method for extracting a spectral EEG feature is provided to be effectively applied to virtual reality by accurately extracting intention and thinking of a subject from a measured noise. A nonnegative N-way tensor is generated by converting a multichannel EEG(Electroencephalogram) data measured from a subject about a specific stimulation into time-frequency representation. A size of the nonnegative N-way tensor is reduced by a data selection process using a nearest neighbor method. A basic component(200) of effective EEG is extracted by nonnegative N-way tensor factorization from EEG including a noise measured about the specific stimulation. The basic component of the effective EEG is discriminated by pattern classification.
展开▼