Audio Emotion Detection CNN Model

This project implements a Convolutional Neural Network (CNN) for detecting emotions in audio samples using the RAVDESS and SURREY datasets.

Table of Contents

• Overview
• Datasets
• Requirements
• License

Overview

This CNN model is designed to classify emotions from audio inputs into 8 categories. It leverages the power of deep learning to extract features from audio spectrograms and predict the corresponding emotion.

Datasets

The model is trained on two popular emotional speech datasets:

1. RAVDESS (Ryerson Audio-Visual Database of Emotional Speech and Song):

   • 24 professional actors (12 female, 12 male)
   • 8 emotions: calm, happy, sad, angry, fearful, surprise, disgust and neutral

2. SURREY Audio-Visual Expressed Emotion (SAVEE) Database:

   • 4 male actors
   • 7 emotions: anger, disgust, fear, happiness, sadness, surprise, and neutral

Requirements

• Python
• TensorFlow
• keras
• Librosa
• Pandas
• Numpy
• Matplotlib
• Sk-learn
• joblib

Usage

1. Clone the repository: git clone https://github.com/Mobeen0/Audio_Emotion_Detection_Using_CNN