In light of the challenges associated with effectively leveraging interactive information across channels and spatial dimensions derived from skeleton key point data, a continuous action recognition method based on multi-branch attention and spatiotemporal graph convolutional network was proposed. Firstly, the interactive information between channel attention and spatial dimension is extracted by embedding multi-branch attention. The features obtained by spatial graph convolution are weighted by multi-branch attention, and then the subsequent temporal graph convolution operation is executed to enhance the representation ability of the model for spatial relations. Finally, the ability of a spatiotemporal graph convolutional network to capture spatiotemporal relationships was used to realize the recognition of continuous actions. The findings from the experiments indicate that the proposed model outperforms current methods in terms of recognition accuracy on the NTU-RGBD and Kinetics datasets.