/Speaker-Recognition

Speaker recognition app using machine learning with recording and import features with GUI.

Primary LanguagePython

Speaker Recognizer

This project is a simple speaker recognition application that uses machine learning to identify speakers from audio recordings. The application is built with Python and utilizes various libraries for audio processing and visualization.

Features

  • Record audio and identify the speaker.
  • Import WAV files and identify the speaker.
  • Visualize audio spectrum during recording.

Requirements

  • Python 3.7 or higher.
  • Required libraries: numpy, matplotlib, librosa, scikit-learn, sounddevice, pyaudio, tkinter, scipy

Installation

  1. Clone the repository:

    git clone https://github.com/yourusername/speaker-recognizer.git
cd speaker-recognizer

  2. Install the required Python libraries:

    pip install -r requirements.txt

  3. Ensure you have the following files in the project directory:

    • microphone.ico
    • background.png
    • background2.png
    • img0.png
    • img1.png
    • img2.png

  4. Create a directory named Voices in the project directory. This folder should contain WAV files for training the model. The filenames should start with the speaker's name (e.g., alice_01.wav, bob_02.wav).

Usage

  1. Run the Application:

    python main.py

  2. Main Window:

    • Click on the microphone icon to start recording.
    • Click on the upload icon to import a WAV file.

  3. Recording Interface:

    • Recording starts automatically after a 1-second delay.
    • The predicted speaker is displayed after recording.
    • The audio spectrum is visualized in real-time.

Code Overview

  • main.py: Main application script.
  • Voices/: Directory containing training audio files.
  • extract_features(): Function to extract features from audio files.
  • import_wav(): Function to import a WAV file and identify the speaker.
  • change_colors(): Function to switch to the recording interface and start recording.

Training the Model

The model is trained using the K-Nearest Neighbors (KNN) algorithm. Training data is loaded from the Voices directory. MFCC features are extracted from the audio files for training.

Example Training Data Structure

Voices/
├── alice_01.wav
├── alice_02.wav
├── bob_01.wav
└── bob_02.wav

Acknowledgements

This project uses the following open-source libraries: