Research on the Development and Application of AI-assisted Teaching Robot System Based on Deep Neural Network

Abstract

This thesis describes an AI-assisted teaching robot system based on a deep neural network, which aims to record the teaching situation in the classroom and continuously improve the teaching method and the quality of education through data analysis. We used the AdaBoost algorithm to implement face detection while using SVM for face classification and recognition. The AdaBoost and SVM algorithms were studied in depth and their effectiveness was verified through experiments. The image acquisition module uses the V4L2 framework to transmit the acquired image data to the client via a wired network, and the network connection adopts the socket network communication mechanism under the Linux system. We also built an Open CV development environment for training the student identification model, realizing student identification, recording the identification, and saving the image data. Based on classroom student behavior detection, this paper achieves more than 90% accuracy by analyzing students' positional features and feature similarity between consecutive video frames. To have a comprehensive understanding of student listening, we quantified students' classroom behaviors using statistical analysis methods, including overall behavioral percentage, behavioral percentage changes, and quantifying students' behaviors into the degree of listening attentiveness by marking scores, to help teachers improve the quality of teaching. This system combines various technologies such as deep learning, image processing, and statistical analysis to bring promising innovations to the field of education.