Publication Details

Abstract

Emotion detection plays a vital role in human–computer interaction, affective computing, and intelligent decision-support systems [1], [7]. Traditional emotion recognition models often struggle with feature redundancy, suboptimal attention mechanisms, and limited generalization across diverse datasets [5], [6]. To address these challenges, this project proposes a High-Precision Emotion Detection Framework utilizing Meta-Heuristic Optimized Twin Attentional Networks.
The proposed framework employs a twin attentional network architecture that simultaneously captures complementary emotional cues from input data, such as textual semantics, facial expressions, or multimodal features [5]. A meta-heuristic optimization algorithm is integrated to automatically fine-tune critical model parameters, attention weights, and feature selection [14], thereby enhancing learning efficiency and classification accuracy.
This optimization-driven approach reduces overfitting, improves convergence speed, and ensures robustness against noisy and imbalanced data [13], [29]. Extensive experiments conducted on benchmark emotion datasets demonstrate that the proposed framework outperforms conventional deep learning and single attention-based models in terms of precision, recall, and overall accuracy [31], [33].