A non-linear method for estimating the alpha generators from an EEG over the scalp.

A method is developed to estimate the frequency feature of alpha generators from EEG signals of different derivation. This method is based on principal component analysis and a constrained non-linear optimization technique. Results of the method gives a variation range of the power spectrum. Experimental results for estimating the generator spectra of the brain alpha activity in anterior and posterior regions over the scalp are shown. Finally, the noise problem with this method is also discussed.