A computer method for identifying patterns in electroencephalogram signals.

A computer method is developed for identifying patterns in electroencephalogram (EEG) signals. An EEG numerical signal is transformed into a symbolic series. The simple transformation used here studies the variations between two successive values of the signal. Then, this series is analysed with a symbolic correlation function based on probabilities without bias. The use of large windows, e.g. 1 hour, allows the identification of weak signals hidden by the specific ones. An application of this method to the sleep analysis of a healthy adult shows a periodicity modulo 10 in all derivations. A possible neurophysiological meaning is presented in the discussion.