[Automatic sleep analysis--II. Determination of sleep stages].

The presented automatic sleep analysis is based on the time courses of the mean amplitudes of EEG (delta, beta and alpha bands), EMG and motility and specific patterns (spindles, REMs, delta waves greater than 40 microV) detected automatically. The interindividual variability of the parameters is taken into consideration by interactively determining the threshold levels. Sleep stages are classified for every epoch using context free decision rules according to the manual of Rechtschaffen and Kales. The resulting cyclograms are smoothed and corrected using context sensitive rules which considers the successive epochs and the probability of stage transitions.