Non-human primates (NHPs) play a critical role in biomedical research. Automated monitoring and analysis of NHP’s behaviors through the surveillance video can greatly support the NHP-related studies. There are two challenges in analyzing the NHP’s surveillance video: the NHP’s behaviors can be seen as coming from an open, possibly incremental set of classes during long-term monitoring, and serious occlusions are brought by the fences of the cages.
In this paper, a feature set combining local sub-block histograms of oriented optical flow (SHOOF) is designed to overcome the effects of occlusions. And based on the proposed feature set, the sticky hierarchical Dirichlet process hidden Markov model (HDP-HMM) is extended to a batch recursive version for jointly segmenting and classifying the NHP’s behaviors. Experimental results on the NHPs’ surveillance video data show significant accuracy in behavior classification, time segmentation and determination of the number of behavior classes.