With the rapid development of information technology, human action recognition technology has gradually evolved from single visual analysis to a new stage of multimodal data fusion. Especially in the field of video action recognition, despite the remarkable progress, it still faces a series of challenges. For example, complex backgrounds, illumination changes, occlusion problems, and viewpoint transformations may affect the recognition accuracy. To address these issues, this paper proposes an innovative framework that combines CSI with video intelligence analysis (yowo) to optimize the human action recognition process. CSI is able to capture small changes in the environment, and by analyzing how the signals change due to the presence and movement of a person, and by utilizing the LSTM + attention mechanism for training and extracting features, it can effectively complement the video analysis The shortcomings in the video analysis, and then the two modalities are fused at a later stage, which not only enhances the system’s ability to adapt to changes in complex environments, but also can alleviate to a certain extent the problems of high false alarm rate and low recognition rate caused by a single reliance on video analysis. This approach is expected to bring new breakthroughs and development opportunities in a number of application areas such as intelligent surveillance, human-computer interaction, and health monitoring.