Advanced Emotion Transformer Networks with Honey Bee Optimization for Identifying Facial Expressions

Facial Emotion Extraction: Facial expressions play a crucial role in informal communication, as they significantly enhance the representation of people's emotions. Identifying human facial expressions using a classification approach has some limitations, such as low accuracy in recognizing exact facial expressions of human beings in classification and complex differences in facial objects, including the eye, nose, and mouth, in segmentation. To solve the issue, propose Emotion Transformer Networks (ETN) with Honey Bee Optimization (HBO) algorithm for the Classification of Student Facial expressions. Initially, the input data set is obtained from the Image Emotion-Net platform, and various facial expressions from multiple student datasets are processed. The Denoising Diffusion Models (DDM) approach identifies a pixel range and removes the blurred area in the input dataset. Next, BiSe-Net V2 is differentiated for specific part segmentation and variation in critical facial regions. A Multimodal Feature Fusion Strategy (MFFS) feature extraction evaluates geometric and presence features to produce more comprehensive information. These features are then used in ETN, which utilizes self-attention mechanisms to capture spatial dependencies between facial components. These features are applied in ETN-HPO with a hyper capsule CNN classification algorithm, which leverages self-attention mechanisms to capture spatial dependencies between facial components. The HBO algorithm is applied to optimize feature selection and hyperparameter evaluation, thereby enhancing classification performance. Finally, the classification output identifies facial expressions: Bored, Confused, Engaged, Drowsy, looking away, the final classification output identifies facial expressions. Moreover, the proposed ETN-HBO approach achieves a sensitivity of 95.9%, a specificity of 95.1%, a precision of 96.0%, an FIScore of 96.1%, and an accuracy of 96.8% metrics. The experimental results show that the ETN-HBO classification significantly improves the detection of complex student facial expressions in real-time applications.