A Hybrid CNN-LSTM Model for Video-Based Teaching Style Evaluation

Intelligent teaching style evaluation has grown in popularity. The video is one of the most accessible resources for evaluation. Yet, there is currently no comprehensive, highly accurate method for processing teaching videos. Thankfully, there are numerous top-notch deep learning-based video categorization methods available as references. A hybrid CNN-LSTM model that is effective at modeling time-series teaching style information is what we propose, drawing on the benefits of video classification research. Our efforts' effectiveness is demonstrated through experiments on a dataset we created for in-class teaching.

I. Introduction

Intelligent in-class quality evaluation has emerged as a result of the growing use of artificial intelligence in education [1]–[4]. The way that teaching quality is typically evaluated is on a scale [5]. However, the majority of the data obtained require time-intensive manual collation and analysis. Automatically recording and intelligently analyzing the teaching video is made possible by intelligent education [6]. Artificial intelligence may replace routine jobs in the classroom, so it's important to assess a teacher's effectiveness using their methods, teaching style, and emotional engagement.