Analysis of Student Facial Reactions During Classroom Instruction

This study seeks to construct a model that is capable of categorizing students' facial expressions during classroom instructional sessions. Facial expressions serve as pivotal non-verbal indicators, offering valuable insights into students' emotional and cognitive conditions. Leveraging Convolutional Neural Network (CNN) technology, this research integrates image models with facial geometry feature models to identify a range of facial expressions including happiness, sadness, neutrality, surprise, anger, and fatigue. The dataset utilized in this investigation comprises 2,069 facial images sorted into six distinct categories of facial expressions. Data collection was carried out through the recording of primary school educational sessions and monitoring of classroom footage sourced from YouTube. Subsequent to preliminary processing steps, encompassing facial landmark recognition, geometry feature extraction, normalization, and data enhancement, the dataset was partitioned into training, validation, and testing sets utilizing an 80:10:10 ratio. The constructed CNN model incorporates six convolutional layers for facial images and four fully connected layers for facial geometry features. Model assessment was conducted utilizing metrics including accuracy, precision, recall, and F1-score. Results from testing reveal that the proposed model exhibits considerable accuracy in recognizing students' facial expressions, with a validation accuracy rate of 92% and a testing accuracy rate of 95%. This research makes a notable contribution to the fields of education and psychology by furnishing valuable feedback on students' emotional and cognitive involvement, thereby supporting educators in pinpointing learning impediments and enhancing instructional methods.