Action recognition method based on depth image and skeleton information

The invention discloses a human body action recognition method based on a depth image and the skeleton information. The method comprises the steps of collecting a human body action video, and extracting the depth image and a skeleton action frame sequence; extracting a global point cloud feature vector from the depth image; generating a skeleton point feature vector from the skeleton action frame sequence; inputting the global point cloud feature vector and the skeleton point feature vector into two SVM classifiers for training to obtain two action classification models; and identifying the category of the to-be-identified human body action sample by using the two action classification models. According to the human body action recognition method based on the depth image and the skeleton information, the motion historical point cloud is generated based on the depth image, the global features of the motion historical point cloud are extracted, the skeleton point features are fused, and the robustness of action