EEG transient detection by matched inverse digital filtering.

Abstract

A computer technique for detection of EEG transients or non-stationarities is described. The procedure entails obtaining the amplitude spectrum (as the square root of the power density spectrum) of a sample of the EEG, in which the non-stationarities themselves have not necessarily been excluded. An inverse digital filter is then formed by inverse Fourier transformation of the inverted smoothed amplitude spectrum, through which the same EEG is then passed. The occurrence of non-stationarities is then signalled if the output of the inverse filter exceeds a preset threshold.