Online Teaching Gestures Recognition Model Based on Deep Learning

摘要

The development of deep learning technology has promoted the application of artificial intelligence in the field of education, and teaching gestures are widely used by teachers in the teaching process. In order to further analyze how teachers' gestures can improve teaching effects and quality, and to make up for the lack of teaching gestures image recognition function and action anthropomorphism, which are currently lacking in educational robots, it is necessary to analyze the different gestures used by teachers in teaching. The key point is how scientific and efficient to recognize all kinds of gestures used by the teachers. Based on the unique convolutional layer and pooling layer structure of convolutional neural network, this paper analyzed five types of online teaching gestures images marked by the human key points detection, including indicative gestures, one-hand beat gestures, two-hand beat gestures, frontal habitual gestures, and lateral habitual gestures. An online teaching gestures recognition model with a ten-layer convolutional neural network is established. The error of this model is minimized by error back propagation and gradient descent algorithms, and the corresponding model parameters are obtained. The model is verified by the test set, and the training accuracy of the model is as high as 96%. The model can be applied to analyze the influence of the teachers' teaching gestures on the teaching effect and quality, at the same time, it can improve the intelligence and anthropomorphism of educational robots in the field of computer vision and its human-computer interaction.