Subspace based spectral subtraction approach for Visual Evoked Potential extraction

This paper presents the extraction of Visual Evoked Potential (VEP) signal from a noisy observation Electroencephalogram (EEG) signal. Subspace method has been used to extract VEP, mostly in the time domain. The author proposes to use subspace based method in the frequency domain for VEP signal estimation. In making use of the subspace method, spectral subtraction approach is utilized. Noise estimate is performed by subspace method in the time domain and subtracts the estimated noise in the frequency domain. Magnitude spectral subtraction is chosen to subtract the noise from the noisy signal. The results show that although the method needs to be developed further, this method can be used in the estimation of VEP signal.