A method and device for recognizing human body movements

Abstract

The present invention is suitable for the field of mode recognition technology. It provides a method and device for human movement, which includes: obtain deep image sequences, transform the depth image sequence to obtain the corresponding depth movement sequence; for what.The division of the time dimension and the division of the space dimension in the depth exercise sequence is obtained by multiple sports historical cubes and the corresponding multiple space sub -cubic blocks; according to the space sub -cubic block calculation, the characteristic vector corresponding to the cube corresponding to the motion history cube is calculated, Combining the characteristic vector corresponding to multiple sports historic cubes, obtain the feature vector of the depth exercise sequence; according to the feature vector of the depth motion sequence, use the support vector machine SVM for model training and testing to obtain human movements to obtain human movementsRecognize the results.The invention solves the problem of self -obstruction of existing human action recognition methods, enhances the ability to describe human movement, and improves the accuracy of recognition of human movements.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition, and in particular relates to a method and device for recognizing human body movements. Background technique [0002] In order to accurately measure the similarity of action tags, a variety of human action recognition methods based on depth image sequences have been proposed in the prior art, such as 3D point clouds, human skeleton models, and hypersurface normal vectors. However, the prior art methods for human action recognition through depth image sequences still have the following disadvantages: [0003] 1. In the case of self-occlusion of the human body, such as "waving hands in front of the chest", the accuracy of action recognition is not high; [0004] 2. The ability to describe the motion information of objects on the time scale is not strong enough, such as "putting things down" and "picking things up"; [0005] 3. The selected features are relatively complex and the data d...

AI Technical Summary

Problems solved by technology

[0006] In view of this, an embodiment of the present invention provides a human motion recognition method and device to solve the problem of self-occlusion in existing human motion recognition methods, enhance the ability to describe human motion, and improve the recognition accuracy of human motion

Images