Abstract
An estimate of the energy spectrum of an evoked potential signal can be based on the spectral density of the averaged evoked potential and the spectral density of the individual evoked potentials. In this paper, the best means of estimating these latter two spectra, based on a fixed data set (fixed number of individual evoked potentials) is considered. Spectral windowing is minimized, in order to avoid temporal distortions caused by the windowing. Results of a simulation study are presented; signal and noise parameters were selected to be similar to the situation of auditory brainstem response recordings. It is concluded that the best estimate of the averaged evoked potential spectral density is based on the Fourier transform of the averaged potential formed from all the data. An adequate estimate of the density of the individual responses can be formed by averaging the Fourier transform of 2 to 8 individual potentials. Finally, some typical signal spectra from human ABR's are presented.