🎤 Speech Emotion Recognition

A machine learning and deep learning-based system for recognizing emotions from speech using audio features like MFCCs, Spectrograms, and more.

📌 Overview

This project implements a Speech Emotion Recognition (SER) pipeline that uses audio signal processing and classification algorithms to detect emotions from speech. It supports multiple datasets, feature extractors, classifiers, and evaluation metrics.

🧠 Supported Emotions

• Neutral
• Calm
• Happy
• Sad
• Angry
• Fearful
• Disgust
• Pleasant Surprise
• Boredom

🛠️ Features

• 🔉 Extracts audio features (MFCC, Chromagram, Spectrogram, etc.)
• 🤖 Classifiers: SVC, RandomForest, GradientBoosting, KNeighbors, MLP, RNN
• 🧪 Hyperparameter tuning via GridSearchCV
• 📊 Evaluation: Accuracy, Confusion Matrix
• 💾 Model saving & loading (.pkl)
• 🔍 Dataset support: RAVDESS, TESS, EMO-DB, Custom

📦 Tech Stack

| Domain | Tools | |--------|-------| | Programming | Python | | Audio Processing | Librosa, OpenSMILE | | Machine Learning | Scikit-learn | | Deep Learning | PyTorch, HuggingFace Transformers (Wav2Vec2) | | Deployment (Optional) | Firebase Functions, Streamlit, Gradio |

📁 Project Structure

bash speech-emotion-recognition/ ├── data/ # Raw and processed audio files, organized by dataset │ ├── RAVDESS/ │ ├── TESS/ │ ├── CREMA-D/ │ └── custom/ # Your own audio recordings │ ├── models/ # Trained models & preprocessed data │ ├── final_model.pkl │ ├── scaler.pkl │ ├── label_encoder.pkl │ ├── tess-model.pkl │ └── tess-label-encoder.pkl # Any .joblib or .pt files │ ├── results/ # Visual outputs │ ├── confusion_matrix.png │ ├── model_accuracy_comparison.png │ ├── src/ # Source code │ └──features.py # Feature extraction scripts ├── train_final_model.py # Training & evaluation logic ├── app.py # Streamlit-based app to demo emotion predictions ├── .gitattributes # Optional: Git LFS or text encoding rules ├── CITATION.cff # Software citation metadata (you-only version) ├── LICENSE # MIT License (under your name) ├── README.md # Main project overview and usage ├── requirements.txt # Python dependencies ├── streamlit_app.py # App interface for demo/testing ├── plot_benchmarks.py # Script to generate accuracy and confusion matrix plots

🚀 Getting Started

1. Clone the repo

git clone https://github.com/chirgamiyy/speech-emotion-recognition.git cd speech-emotion-recognition

1. Install dependencies

pip install -r requirements.txt

1. Run training or prediction

python src/train.py # Train model python src/predict.py # Predict emotion from audio

📊 Example Results

🔹 Model Accuracy Comparison (93.96%)

🔹 Confusion Matrix (on Combined Dataset)

📜 License

This project is licensed under the MIT License.

🙌 Acknowledgements

• Audio Datasets:

  • RAVDESS on Kaggle
  • TESS on Kaggle
  • CREMA-D on Kaggle

• Feature Extraction Libraries:

  • Librosa
  • OpenSMILE Toolkit

• Machine Learning & Deep Learning:

  • Scikit-learn
  • PyTorch
  • HuggingFace Transformers

If you build upon this work, please consider citing it via the CITATION.cff file.

📚 Citation

If you use this work, please cite it using the metadata in CITATION.cff.

bibtex @software{agrawal_2025_ser, author = {Chirag Agrawal}, title = {Speech Emotion Recognition}, year = {2025}, version = {1.0.0}, url = {https://github.com/chirgamiyy/speech-emotion-recognition} } Feel free to ⭐ the repo if you found it helpful!