Voice Authentication System in Python

A simple voice-based authentication system built with Python. Record a voice password, train a machine learning model using MFCC features, and run real-time voice validation to grant or deny access.

Inspired by spy-movie-style biometric systems — but built entirely with open-source tools, running locally.

---

Features

• Record your own voice password
• Extract MFCC features with librosa
• Train an SVM model using scikit-learn
• Real-time voice authentication from mic input
• Lightweight, local-only — no cloud APIs

---

How It Works

1. Record audio samples of your password phrase (e.g., "access")
2. Record negative samples (e.g., saying a different word like "banana")
3. Extract features from the audio files using MFCC
4. Train a classifier to distinguish between your voice and others
5. Use the model to authenticate future voice input

---

Project Structure

auth-voice-system/
├── authenticate.py
├── extract_features.py
├── record_voice_password.py
├── train_model.py

Note: The voice samples, extracted features, and trained model are not included for privacy reasons. Run the scripts yourself to generate them.

---

Getting Started

1. Install Requirements

pip install sounddevice scipy librosa scikit-learn numpy joblib

2. Record Voice Samples

Record 20 samples of your password phrase:

python3 record_voice_password.py

Then change the phrase (e.g., to "banana") and record 10 negative samples into a different folder.

3. Extract MFCC Features

python3 extract_features.py

This generates a dataset from your recordings.

4. Train the Model

python3 train_model.py

This saves voice_auth_model.pkl.

5. Run Authentication

python3 authenticate.py

Speak your password — the system will predict if it's you and decide whether to grant access.

---

What You Can Build With This

• Raspberry Pi door unlocker
• Secure local script runner
• Biometric voice login for private folders
• Smart home voice toggle
• Multispeaker verification system

---

What to Try Next

• Use deep learning (e.g., CNNs or speechbrain) for better accuracy
• Add a simple GUI with Streamlit or Flask
• Hook it to real hardware using GPIO (e.g., Raspberry Pi relay)
• Add noise filtering or augmentation for robustness
• Detect spoofing attempts with audio playback detection

---

License

This project is open-sourced for educational purposes. Use responsibly. No biometric data is included in this repository.