A Facial Feature Recognition-based System for Evaluation of Teaching Quality in College English Classrooms

The quick advancement of big data technology has contributed to a greater importance of online learning, thus particularly in the way of online courses. A face feature recognition-based approach is proposed for evaluating the quality of teaching in English classrooms. An Improved Cat Swarm Optimization (ICSO) approach is proposed to tuning the hyperparameter to adjust DenseNet 121 for accurately determining whether the mouth and eyes are open or closed. For increasing recognition accuracy, Attention Mechanism (AM) is used with DenseNet121 to avoid over-fitting issue. Due to its compact size and efficient architecture, the over fitting issue in the channel attention network also avoided. Lastly, the learning focus of the learners is assessed using the fuzzy comprehensive assessment approach. The findings of the simulation studies demonstrate that the DenseNet121-AM-ICSO outperforms the conventional approaches such as Multitask Convolutional Neural Network (MT-CNN) and Spatial AM with Residual Network (Sd_ResNeSt) a higher recognition accuracy of 99.79%.