An Advanced Framework for Emotion Recognition from Speech using Deep Convolution Approach

Emotions are crucial for understanding a person's feelings and mental state. They provide a way to express thoughts and feelings to others. The aim of this research is to examine speech signals for the extraction of emotions, contributing valuable insights into understanding an individual's mental state. Essential features such as tone, pitch, and energy are crucial components for training a classifier model that accurately identifies emotions. To compile the necessary data for this study, four distinct datasets—SAVEE dataset, RAVDESS dataset, TESS dataset, and CREMA-D dataset—are employed. The development of neural networks and the increasing need for precise SER in human computer interaction requires a comparison of methods and databases. This helps in finding practical solutions and better understanding the problem. The study analysis deep learning and traditional machine learning techniques for SER, comparing applied neural network perspective for SER. In summary, this research underscores the significance of feature extraction from speech signals and the diverse applications of different datasets and machine learning. The emphasis on the superiority of the 2D CNN model underscores its suitability for effective emotion classification in speech signals.