Research On Dance Student Posture Estimation In Intelligent Classroom Scenarios

Dance student posture estimation in smart classrooms can well reflect students’ classroom state,reflect their mastery of learning content,and is of great significance to the digitization of educational processes.In traditional dance classes,teachers judge the standard level of a dancer’s dance by observing their students’ body movements,but sometimes they cannot observe whether the movements are standard in time,and they also repeatedly observe the students’ dance movements to discover the characteristics of their dance.Therefore,it is particularly important to apply artificial intelligence technology to evaluate the advantages and disadvantages of dance movements.This article aims at solving the problem of time-consuming and labor-intensive manual observation of the posture of dancers in intelligent classrooms,and uses convolutional neural networks to detect and extract human joint points to predict the posture of dancers in intelligent classroom scenes.It proposes a posture estimation method for intelligent classroom students based on human skeleton and deep learning,which can easily and effectively reflect the dance movements of dancers and help teachers accurately observe the students’ dance posture standards,It is conducive to improving the quality of dance in smart classrooms.The main research contents of this paper are as follows:(1)Construct the pose estimation data set of intelligent classroom dancers.Currently,no published human posture estimation data set exists in the wisdom classroom scenario.In order to achieve better application effect,this paper downloads a large amount of data on various dance movements in different dance classrooms online.Finally,4000 images were selected to construct the Intelligence Classroom Dancer Posture Estimation Database(hereafter referred to as ICDPED).(2)Design a posture estimation method for intelligent classroom dancers based on deep learning.Detect the availability of the dataset.Research the relevant theoretical knowledge of human posture estimation technology,introduce two classic human posture estimation models,Resnet and Higher HRNet,and conduct comparative experiments on these two network models.The accuracy rates are 72.6% and 70.7%,respectively.Therefore,Higher HRNet network is selected to extract human key points and predict the posture of intelligent classroom dancers.(3)Introducing convolutional module attention mechanism(CBAM)and empty space pyramid pooling(ASPP)into the Higher HRNet network to optimize the posture estimation network for intelligent classroom dancers.The CBAM attention mechanism improves the extraction rate of human feature information in both channel and spatial aspects of the feature network,with an accuracy rate of 1.5%;ASPP can expand the receptive field and capture more feature information through multi-scale feature fusion,thereby improving the accuracy of the network.Improving the accuracy of student posture estimation based on improved network structure 2.6% higher.It is further proved that the model obtained from this experiment can be applied to intelligent classrooms,reducing the workload of teachers,and improving the quality of classroom dance.