On the Sufficiency of Autocorrelation Functions as EEG Descriptors

Brain waves or EEG´s are the seemingly random voltage fluctuations which appear on the surface of the scalp of humans and animals. The EEG Research Group at the University of Missouri has been interested in finding useful statistics with which to describe an EEG. In the past such calculations as the rms value of the wave, the autocorrelation function, etc., have been used to describe the process. It is convenient to use a stationary random process as a model for the EEG. Because the amplitude distribution of the EEG appears to be Gaussian, it has been suggested that a better model might be the normal stationary random process. This paper describes a test which was made to determine if the normal process is really a good model for the EEG.