Аннотация:Video tracking systems are widely used for behavior analysis of laboratory animals. They perform tracking of 2-3 typical points of animals. For further analysis it is required to identify separate behavioral acts. Currently this can be done only manually by a specialist in animal behavior. Existing automatic methods are able to mark the periods of immobility and active motion. They have relatively low accuracy and require careful parameter setting. In the paper we suggest video tracking system which models animal behavior by using hidden Markov models. The system shows promising results in analyzing mouse behavior in the new environment.