Exploration of different audio features and CNN-based architectures for building an effective Speech Emotion Recognition (SER) system. The goal is to improve the accuracy of detecting emotions embedded in speech signals. The repository contains code, notebook, and detailed explanations of the experiments conducted.

Audio system using 1D CNN and 2D parallel CNN and CNN using Transformer By studying numerous vocal and speech patterns, this project proposes an automated technique to recognizing human emotions.

This project implements a dual-mode Emotion Recognition system using deep learning. It trains separate models for: Facial emotion recognition using the FER-2013 dataset Audio emotion recognition using the RAVDESS dataset Both training scripts generate h5 and png. Deployment is not yet integrated; current focus is on training and model generation.