Research Progress on the Human Behavior Recognition Based on Machine Learning Methods

Machine vision has been used in the industrial automation system for a long time. It also plays a significant role in the field of human behavior recognition. Behavior recognition based on machine vision, such as object tracking, motion detection and crime recognition, greatly broadens the application field of artificial intelligence and has a good application prospect. We summarize the latest applications of various machine learning algorithms in human behavior recognition, and analyze the accuracy of various algorithms combined with data sets, so as to provide reference for researchers in related fields. By sorting out the typical research results, briefly expound on the application of machine learning in the field of behavior recognition in recent years. This review focuses on the Two Stream Network structure, TSN structure, LSTM network and C3D network. This paper analyzes the principles, advantages and disadvantages of various human behavior recognition methods, and briefly discusses the future development direction. The wide application prospect of behavior recognition and detection makes it a hot research direction in the field of computer vision, and greatly improves the accuracy of complex human motion recognition combined with deep learning. However, it still faces many difficulties, such as insufficient discrimination of violence attributes, difficult collection, verification of special action data and insufficient hardware computing resources, etc.