SentimentSound

Overview

This is a deep learning model for Speech Emotion Recognition that can classify audio clips into different emotional states. The model is trained on a dataset of speech samples and can identify emotions such as neutral, calm, happy, sad, angry, fearful, disgust, and surprised.

Model Details

• Model Type: Hybrid Neural Network (CNN + LSTM)
• Input: Audio features extracted from 3-second wav files
• Output: Emotion classification

Supported Emotions

• Neutral
• Calm
• Happy
• Sad
• Angry
• Fearful
• Disgust
• Surprised

Installation

Clone the Repository

git clone https://github.com/Vishal-Padia/SentimentSound.git

Dependencies

pip install -r requirements.txt

Usage Example

python emotion_predictor.py

Model Performance

• Accuracy: 85%
• Evaluation Metrics: Confusion matrix below

Training Details

• Feature Extraction:
  • MFCC
  • Spectral Centroid
  • Chroma Features
  • Spectral Contrast
  • Zero Crossing Rate
  • Spectral Rolloff
• Augmentation: Random noise and scaling applied
• Training Techniques:
  • Class weighted loss
  • AdamW optimizer
  • Learning rate scheduling
  • Gradient clipping

Limitations

• Works best with clear speech recordings
• Optimized for 3-second audio clips
• Performance may vary with different audio sources

Acknowledgments

• Dataset used for training (https://www.kaggle.com/datasets/uwrfkaggler/ravdess-emotional-speech-audio)