A Human Behavior Recognition Method Based on Self-feedback Gene Expression Programming

Abstract

The invention discloses a human body behavior recognition method based on self-feedback gene expression programming. The method is aimed at the depth image of human body behavior, extracts three-dimensional time series data of multiple joint points of the human body as a sample, and uses the gene expression programming in After the crossover and mutation operations, add the TIS interpolation operation to construct the self-feedback gene expression programming to model the samples and obtain the human motion model. The TIS interpolation operation refers to inserting a function string at the head of the joint motion sequence; then Gradient information is extracted as model features. Input the model features of the human body motion model of the training sample into the neural network, and train the neural network model as a human body behavior classifier; input the model features of the human body motion model corresponding to the test sample into the human body behavior classifier obtained above, and obtain the human body behavior classifier. behavior recognition results. The invention has the advantages of high recognition accuracy and fast recognition speed of human behavior.

Description

technical field [0001] The invention relates to the field of human behavior recognition, in particular to a human behavior recognition method based on self-feedback gene expression programming. Background technique [0002] In recent years, human behavior recognition research is one of the most popular frontier research topics in computer science in the world. It has important theoretical research value, and the research topics are rich in content, involving image processing and analysis, machine vision, human physiology, and human kinematics. , pattern recognition and artificial intelligence and other multidisciplinary domain knowledge. The research on human behavior recognition not only broadens the research direction between disciplines, but also drives the development of related disciplines. It is also a key technology that urgently needs to be solved in smart cities and public safety. It is widely used in human-computer intelligent interaction, intelligent visual monito...

AI Technical Summary

Problems solved by technology

At present, most of the research on human body recognition has only stopped at static recognition. The complexity and variability of human body movement make the accuracy and efficiency of recognition unable to meet the practical requirements of related industries.

[0003] At present, the recognition of human behavior in the prior art mainly uses joint-point wearing sensors or multi-cameras for multi-view monitoring. Joint-point wearing sensors refer to placing sensors on a certain part of the human body. The types of behaviors are limited, and the recognition accuracy is not high, the recognition speed is slow, the recognition behavior is single, and the scalability is poor; the use of multiple cameras for multi-view surveillance refers to installing cameras at various viewing angles, and judging the human body through the pictures captured by the cameras of each viewing angle. This method has defects such as high cost, low data accuracy, inconvenient use, and great influence by environmental factors.

Images