A human behavior recognition method based on hierarchical feature subspace learning

Abstract

The invention discloses a human behavior recognition method based on hierarchical feature subspace learning. The method comprises: dividing a human behavior video sample into a training sample and a test sample; extracting and encoding the manual feature and depth feature of the video sample; respectively representing each video sample with the feature vector of the manual feature and the depth feature; Setting sample labels of video samples, using manual feature and depth feature of training samples and corresponding sample labels as inputs, training hierarchical feature subspace learning model, generating subspace projection matrix and decision boundaries of behavior categories; Learning the subspace eigenvectors of the manual and depth features of the test samples by using the subspaceprojection matrix; Calculating and comparing the distance between the subspace eigenvector of the test sample and the decision boundary of each behavior class, and discriminateing the behavior class of the test sample to complete the recognition operation; The invention improves the identification power of the subspace feature expression, and has good identification accuracy and efficiency.

Description

technical field [0001] The invention belongs to the technical field of video behavior recognition, and in particular relates to a human behavior recognition method based on hierarchical feature subspace learning. Background technique [0002] Video-based behavior recognition is a research hotspot in the field of artificial intelligence development and computer vision. It has important market demand and application value in the fields of intelligent security monitoring, intelligent robots, human-computer interaction, virtual reality and game control. Learning video feature expressions with strong discrimination is the key to improving the accuracy of action recognition. However, the similarity of different actions and the difference of the same action increase the complexity of action description and recognition. [0003] At present, most scholars are committed to mining new manual low-level features, or building deeper and more complex deep learning models, and exploring fea...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a human behavior recognition method based on hierarchical feature subspace learning in view of the problems of complex human behavior recognition models in the prior art, low algorithm efficiency and slow recognition speed. From the perspective of learning, the fusion of complementary feature expressions based on manual features and deep features realizes the recognition operation of human behavior. The specific technical solutions are as follows:

Images