Human body behavior identification method based on residual-recurrent neural network

The invention provides a human body behavior identification method based on a residual-recurrent neural network. The human body behavior identification method comprises the following steps: acquiringinformation of a plurality of human body skeleton key points; obtaining a training set, a verification set and a test set; constructing a residual error-recurrent neural network; training the residualerror-recurrent neural network; and obtaining a human body behavior recognition result. According to the invention, the single-person image in the video is extracted through detection; skeleton key point detection is carried out to obtain human skeleton key points; the residual-recurrent neural network is utilized to realize final behavior identification, so space and time dimension characteristics are fused on the basis of removing redundant information in the video, model calculation amount is greatly reduced, the extracted characteristic information is more comprehensive, and accuracy andspeed of human body behavio