Human-object Behavior Analysis Based on Interaction Feature Generation Algorithm

Aiming at the problem of insufficient utilization of interactive feature information between human and object, this paper proposes a two-stream human-object behavior analysis network based on interaction feature generation algorithm. The network extracts human-object’s feature information and interactive feature information respectively. When extracting human-object features information, considering that ResNeXt has powerful feature expression ability, the network is used to extract human-object features from images. When extracting interactive features information between human and object, an interaction feature generation algorithm is proposed, which uses the feature reasoning ability of graph convolutional neural networks. A graph model is constructed by taking human and objects as nodes and the interaction between them as edges. According to the interactive feature generation algorithm, the graph model is updated by traversing nodes, and new interactive features are generated during this process. Finally, the humans’ and objects’ features information and the human-object interaction feature information are fused and sent to the classification network for behavior recognition, so as to fully utilize the humans’ and objects’ feature information and the interaction feature information of human-objects. The human-object behavior analysis network is experimentally verified. The results show that the accuracy of the network has been significantly improved on HICO-DET and V-COCO datasets.

@article{Ye2023,
  title     = {Human-object Behavior Analysis Based on Interaction Feature Generation Algorithm},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {14},
  number    = {8},
  year      = {2023},
  publisher = {The Science and Information Organization},
  author    = {Qing Ye and Xiuju Xu and Rui Li},
  doi       = {10.14569/IJACSA.2023.0140859},
  url       = {https://doi.org/10.14569/IJACSA.2023.0140859}
}