An optimal monitor of the rapid-eye-movement brain state

A model has been proposed for the stochastic occurrence of bursts of rapid eye movements (REMs) during sleep. REM-burst are simulated by a Poisson counting process with a rate that depends on a binary Markov "sleep state". The corresponding maximum likelihood detector, that continuously monitors the current sleep state based on the observed REM-bursts, has been derived and implemented. Simulations show 97% correct decisions.