Human Action Recognition And Application In The Education Recording System

The algorithm study of human actions recognition based on videos has become a hot topic in computer vision.In recent years,it has gained more achievements,which applied to video surveillance,road traffic,virtual reality,medical care,sports and other fields.In addition,with the support of education in our country and attaching great importance to multimedia teaching has been widely popular.The teaching process recorded is one of the main developments,so the intelligent video recording and broadcasting system based on video image tracking and identification technology comes into being.It has changed the teaching mode that teacher teaches on the podium and students sit under listening,which makes students and others can learn more knowledge at any time,to achieve the teaching resources sharing and reusing.Education recording system based on videos mainly includes teachers tracking system and students positioning system,so this paper researches on recognition algorithm of students in classroom to meet the actual demand.This paper mainly studies the method of human action recognition,and analyses recognition algorithms with classroom environment,to recognize students behaviors while speaking or answering questions,mainly to identify three actions including hand,stand and sit.Aiming at the background of the study scene and the characteristics of the student action itself,this paper presents a Zernike moments based on the motion history image and the classification of the Naive Bayesian classifier.And judging the direction of motions through the Lucas-Kanade optical flow characteristics and the global motion direction features to identify actions.In foreground extraction,studing several commonly used moving object detection methods and comparing experimental results.According to the scenario in this paper,using the background subtraction method to detect foreground.In feature extraction,Zernike moments feature are extracted based on motion history image,extracting the direction characteristics of the motion through the optical flow feature and the global motion direction feature.In the classification and recognition,using the Naive Bayesian Classifier to classify the student three actions,and dividing hand,stand and sit two actions for two categories.Based on video database of Student-Behavior captured in the paper to experiment test,the experimental results show that the proposed method can effectively identify these three actions in the complex classroom environment,and can accurately identify the simulated interference scenes with teachers and students,and the recognition rates are high,which has certain feasibility and robustness.