Human skeleton behavior recognition model based on multi-object pose estimation with spatiotemporal semantics

Abstract Multi-object pose estimation in surveillance scenes is challenging and inaccurate due to object motion blur occlusion video data. Targeting at the temporal dependence coherence among frames, this paper reconstructs a multi-object model that integrates spatiotemporal semantics for different scales poses of multi-objects. The firstly, with an end-to-end detection framework, detects multiple targets video. Secondly, it enhances positioning key points human body using cues frames designs modular components enrich information, effectively refining estimation. Finally, improved skeleton behavior recognition based on employed recognize classroom behaviors students oriented streams. Comparison classifiers through experiments reveals combined exhibits accuracy.