New Approach to Multimodal Emotion Recognition Based on Deep Learning and Feature Fusion Techniques

Intelligent systems have transformed public services, yet challenges remain in ensuring their safe and effective operation. Emotion recognition is crucial for enhancing human-computer interaction. This paper presents a multimodal emotion recognition system integrating deep learning techniques. Audio features are extracted from frequency characteristics, while textual data undergo tokenization. A Fusion Network technique is introduced to enhance performance by combining multiple models. Evaluated on the IEMOCAP dataset, our approach achieves 72% accuracy, demonstrating its effectiveness for emotion-aware applications.