Graph Convolutional Skeleton-based Action Recognition Method for Intelligent Behavior Analysis

Smart education is education informatization, which is a new generation of education model using modern information technology. Smart behavior analysis is the core component of the smart education system. In the face of complex classroom application scenarios, the accuracy and There is a serious shortage of timeliness, and a Depthwise Separable Attention Graph Convolutional Network (DSA GCN) skeleton action recognition algorithm based on separation and attention mechanism is proposed. The inherently insufficient problem is to perform multi-dimensional channel mapping through point-by-point convolution, combining the ability of spatiotemporal graph convolution to protect the original spatiotemporal information of the input skeleton sequence with the separation ability of depthwise separable convolution in spatial and channel feature learning. , in order to enhance the model feature learning and abstract expression. Secondly, the multi-dimensional fusion attention mechanism is adopted, and the self-attention an