Research On Student Classroom Behavior Recognition Algorithm Based On Convolutional Neural Network

In recent years,with the rapid development of smart campus construction in China,the deep integration of information technology and education and teaching has become an inevitable trend.In classroom teaching,student behavioral states can reflect student classroom participation and teaching effect,which is of significance for the evaluation of teaching quality and the improvement of teaching methods.At the same time,with the development of artificial intelligence,deep learning has achieved a lot of achievements in many fields such as speech,image recognition,natural language processing,and new framework models are constantly emerging.Among them,convolutional neural network(CNN)has become one of the research hotspots in many scientific fields due to its simple structure,few training parameters,and strong adaptability.This paper studies the target detection algorithm based on convolutional neural network,and applies it to the recognition of student classroom behavior states,which has important research significance and application value.The main research contents are as follows:First of all,this article elaborates the research background of the application of artificial intelligence in the field of education and the research status of deep learning algorithms,and analyzes the application prospects of target detection algorithm in the recognition of student behavioral states.At the same time,this paper introduces the classic models and target detection algorithms of deep learning,and compares and analyzes the performance of the algorithms.Secondly,Aiming at the YOLOv3 algorithm in convolutional neural networks has the problems of low accuracy on small target detection,and missed detection in classroom behavior state of students detection,an improved target detection method based on YOLOv3 algorithm is proposed.On the one hand,the method uses K-means algorithm to perform cluster analysis on the training dataset,so as to obtain the anchor size suitable for the training of the dataset.On the other hand,through the combination of region division of input image and effective target layer extraction,which increases feature extraction of small targets by the deep convolution layer,thereby further improving the detection accuracy of the YOLOv3 algorithm.Experimental results show that the average accuracy(m AP)of the improved YOLOv3 algorithm on the VOC2007 validation set reaches 78.8%.In addition,the recognition effect of student behavior status has been further improved,and the value of m AP has reached 88.6%.Thirdly,Aiming at YOLOv3 algorithm has the problems of imbalance of positive and negative samples in training and weak classification detection ability,and further improves the YOLOv3 algorithm.On the one hand,this method uses the Online Hard Example Mining(OHEM)algorithm to deal with the self-built student classroom behavior dataset,and improves the imbalance of positive and negative samples in training process.On the other hand,the RPN structure is introduced after YOLOv3 feature extraction network layer to further detect and classify feature maps.The experimental results show that the improved method improves the detection accuracy of the algorithm on the self-built student classroom behavior dataset,and also improves the algorithm’s missed detection of multiple behavior states.Among them,the verification value of average accuracy(m AP)reached90.2%.Finally,this paper summarizes the work of this article,and made a further prospect for the application of deep learning in education and teaching.