A modified FFT-based filter for noise reduction of EEG measurements in identifying task learning processes

Summary form only given. A novel digital bandpass filter, which yields noise-free measurements of EEG signals, is presented. The accuracy of the digital bandpass filter depends on the performance of the discrete Fourier transform method, which is adopted in the filter. A revised discrete Fourier transform method is proposed to decrease the transformation error due to the finiteness of the EEG time series and the number of Fourier coefficients. This method requires much more calculation time; however, the transform error of the method is considerably reduced compared with ordinary FFT. The resulting filtered EEG signals are subsequently incorporated in a scheme for identifying task learning processes in humans via EEG analysis