Audio Fingerprinting Project

Note: This is a refactored version from the April 2025 repo

This project implements an audio fingerprinting system combining the constellation mapping, spectrogram peak extraction, hashing techniques, and pre-selection process for robust audio matching. It aims to identify audio files based on their unique spectral patterns, making it useful for tasks such as audio similarity search.

Instructions:

1. Create a virtual environment and activate

 python3 create venv myvenv
 source myvenv/bin/activate

2. Install dependencies:

pip install -r requirements.txt

3. Run the main script for end to end execution using the following command in the terminal:

python3 main.py

Requirements:

1. Make sure you have Python 3.10+ installed.

Importing files (Input Files)

The system takes .wav or .mp3 files as input for fingerprinting. Import the database recordings to the database_recordings directory.

'./data/database_recordings'

Import query files

'./data/query_recordings'

Project Structure

├── main.py                     # Main script for running the app
├── finger_print                # Where the fingerprint hashes are stored
    └── db_audio.pkl
├── notebooks                   # Contains plots and evaluation (F1, Precision and Recall)
    ├── evaluation.ipynb    
    └── plot.ipynb
├── scripts
    ├── __init__.py
    ├── audio_match.py          # Matching algorithm for query audio files
    ├── config.py               # Stored variables 
    ├── fingerprint_build.py    # Algorithm for building the fingerprint
├── requirements.txt            # Python dependencies for the project
├── README.md                   # Instructions
└── data/                       # Folder for storing audio files 
    ├── database_recordings/    # Store database audio files
    └── query_recordings/       # Store query audio files

Output files

Output files are located in output.txt Example:

query: classical.00003-snippet-10-10.wav database: classical.00003.wav pop.00003.wav classical.00017.wav