The paper presents a new algorithm for adaptive classification of sleep stages using multimodal data recorded on the sleep laboratory during overnight polysomnography records.