HThe invention discloses a human body behavior identification method and system based on attention perception and a tree skeleton point structure

Abstract

The invention discloses a human body behavior identification method and system based on attention perception and a tree skeleton point structure. The method comprises the following steps: 1) inputtingskeleton point information of all behavior samples in a training set; 2) enabling the number of frames of each sample to be consistent by adding a zero-padding frame; 3) reconstructing disordered skeleton points by using a three-way tree traversal rule; 4) performing normalization processing on the Laplacian matrix of the feature map obtained after reconstruction; 5) constructing an attention perception network and a main body network; T. The behavior recognition method comprises the steps of (1) establishing a behavior recognition network model, (2) establishing a behavior recognition network model, (6) connecting the attention perception network with the main body network in a hierarchical manner, (7) inputting the reconstructed feature map into the main body network and the attention perception network respectively to train the behavior recognition network model, and (8) carrying out behavior recognition by utilizing the trained behavior recognition network model.

Description

technical field [0001] The invention belongs to the field of robot vision technology and human-computer interaction, and specifically relates to a human behavior recognition method and system based on attention perception and tree-shaped skeleton point structure; better recognition is achieved by tree-shaped reconstruction of disordered skeleton points Describe the expression behavior pattern, and use the attention network to describe the importance of different skeleton points, and provide suitable prior conditions for the classification of human behavior, which can further reduce the classification processing time of human behavior recognition and improve the accuracy of later behavior recognition . Background technique [0002] Behavior recognition belongs to the field of behavior analysis. For a given video sequence containing a certain type of motion, the video sequence is tagged according to the type of motion, which can be applied to human-computer interaction, intell...

AI Technical Summary

Problems solved by technology

However, the network complexity based on the LSTM structure is high. Although the detection accuracy of the network after the attention mechanism has been improved, its algorithm efficiency is low, which is inconvenient for subsequent maintenance and improvement, and is not suitable for application in actual scenarios.

Images