Exploration of Emotion Analysis and Personalized Emotion Adjustment Algorithm in English Learning Process

Abstract

An important component of human existence and behaviour is emotion. The automatic identification of emotions has gained significant importance in the domains of affective computing and human-machine interaction in recent times. Getting pictures of people's facial expressions is one of the simplest and least expensive methods for identifying emotions among the several physiological and kinematic data that may be utilised. Because of variations in anatomy, culture, and environment, developing a generalised, cross-subject model for emotion identification from facial expression remains difficult.This study presents an intelligent classroom that uses e-learning to help persons with autism spectrum condition strengthen their emotional abilities. Monitoring autism child engagement and participation in class while looking for signs of attention is necessary for effective classroom education. Teaching quality is increasingly being determined by the teachers' capacity to assess and evaluate the behaviour of their pupils in the classroom. Modern classroom systems are enhanced with the most recent technologies in educational institutions to make them more interactive, student-centered, and personalised. Even with these tools, it can be challenging for teachers to gauge their pupils' levels of interest and focus. This study makes use of contemporary technology to establish a real-time, intelligent vision-based classroom that can track students' moods, attendance, and levels of focus even when they are wearing face masks. In order to evaluate students' attention or lack thereof in a classroom, we trained AlexNet models that recognise student behaviour, including recognising facial expressions. Another designed to manage the irregular relationship between pupils and activities is the Interaction Students Education Neural Network ( Int_Edu_NN). The Q-matrix is used to firstly determine the pupil variable and the assignment indicator. This network is analyzed using two datasets in terms of various parameters and hence the proposed Int_Edu_NN achieves 99.% and 98% of accuracy.