Speech emotion recognition is pivotal in human-computer interaction, enabling machines to understand and respond to human emotions. While traditional methods often rely on low-level feature extraction, this paper proposes a language-independent, deep learning-based framework for accurate speech emotion classification. By combining the strengths of hybrid feature extraction techniques and 3D Convolution Neural Networks (CNN), the proposed framework effectively captures both local and global information from speech signals. The model is evaluated on RAVDESS, CREMA-D, SAVEE, and TESS datasets, achieving impressive accuracy rates of 98.48%. In order to verify the proposed work, we have compared the model with the LSTM and Bi-LSTM models too. Experimental results demonstrate the superiority of the proposed framework over state-of-the-art methods, paving the way for more sophisticated and empathetic human-computer interactions.
A Thu, study studied this question.