Spectral estimation with a non-parametric multiple-window method

The spectral characters of the electroencephalogram activity (EEG) is of interest both in studies of the EEG and in estimation oI Evoked Potentials. Often parametric methods assume an underlying model like AR or ARMA [1]. To avoid the model assumption a non-parametric multiple-window meihod for spectral estimation is used in this paper. The method uses Slepian's Discrete Prolate Spheroidal Wave Functions and produces estimates with low bias and high resolutlion even when the data length is short