Accurate calibration of physical education actions based on artificial intelligence deep learning technology

Abstract

Allowing computers to recognize human actions and behaviors based on video sequences captured by cameras is an important branch of computer vision and AI. At present, human action recognition has been widely used in security monitoring, interactive entertainment, smart home and other aspects. It "aims at cultivating students' ability", which is an educational mode of online education for students with the help of network information platform before class, focusing on collaborative inquiry, interactive communication, promoting knowledge internalization and consolidating knowledge after class. Inspired by the visual mechanism of human brain, the proposal of deep learning makes a breakthrough in machine learning, and also brings a new direction to the research of human motion recognition. Deep learning is based on a series of algorithms, and unsupervised high-level abstraction of data is obtained through hierarchical nonlinear transformation. This paper focuses on human motion recognition in complex scenes and the extraction of spatio-temporal features in motion video, so as to overcome the difficulties caused by environmental differences and time changes. Based on the study of CNN and deep confidence network, an innovative human motion recognition model is proposed in this paper.