Knowledge distillation and multi-task self-supervised learning-based skeleton behavior identification method

The invention relates to the technical field of skeleton behavior recognition, and relates to a skeleton behavior recognition method based on knowledge distillation and multi-task self-supervised learning, which comprises the following steps: 1) in a data preprocessing stage, dividing an acquired skeleton sequence into two parts through joint point-level body segmentation; 2) in a network pre-training stage, designing a multi-task self-supervised contrast learning framework, extracting feature representation information from a plurality of tasks through a coding network, inputting the feature representation information into cross-task bidirectional knowledge distillation, and adjusting similarity information between samples through a loss function; meanwhile, weight parameters are stored; 3) in a network fine tuning stage, initializing the coding network according to the weight parameters, and constructing an action recognition model in combination with the initialized coding network and the fine tuning class