Freezam: A Shazam-like Audio Recognition Algorithm

Audio fingerprinting and recognition algorithm implemented in Python.

Users may create a customized music database, view spectral analyses of a song, and identify a song from noisy snippets. Main functionality of this program has been tested on Windows 10.

Dependencies

Software

• ffmpeg for converting audio files to .wav format
• PostgreSQL for database construction

Python packages

• pydub a Python ffmpeg wrapper
• eyed3 for reading mp3 metadata
• numpy for audio signals transformations
• scipy used in spectrogram and peak finding algorithms
• matplotlib used for spectrogram plots
• psycopg2 a Python-PostgreSQL database adapter

Installation

First, install the above dependencies.

Second, git clone the project into a local git directory.

Third, you'll allow the program to access your PostgreSQL database where fingerprints can be stored. In the shazam folder, create a python file named credentials.py:

#credentials.py

DB_USER = 'your-db-username'
DB_PASSWORD = your-db-password

Now you're ready to start fingerprinting your audio collection!

Description

This program has the following functionalities:

Database construction

This program allows you to build your own music database at 1-click!

To get started, please copy your music files (preferably in mp3 format) into the freezam/music/mp3 folder. You'll notice that the folder already contains some pre-downloaded music files for testing purposes. Feel free to add or remove files in the folder.

Then run the following command in the terminal:

cd freezam
$python interface.py construct

The program will print a message when it is done.

Database management

Currently the program supports the following manipulations of database:

• add a song to database

python interface.py add [-h] [--pathfile PATHFILE]

• modify song info

python interface.py update [-h] [--title TITLE] [--artist ARTIST] [--album ALBUM]

• remove a song from database

python interface.py remove [-h] [--title TITLE]

• list all songs in database

python interface.py list [-h]

• check and remove duplicate entries (should run regularly for database maintenance)

python interface.py admin --action=rm_dup

• More to come...

Identify a snippet

python interface.py identify [-h] [--pathfile PATHFILE] --type=1

or

python interface.py identify [-h] [--pathfile PATHFILE] --type=2

This program implements two types of fingerprints for audio identification:

• type=1 computes a signature from local periodograms using the peak positive frequency method.
• type=2 computes a signature by finding the maximum power per octave in local periodograms.

For faster identification, choose type=1; for better precision, choose type=2. The default option is type=2.

Logging

This application writes a message for each action taken to a designated log file shazam.log. Warnings and error messages go to the log file but also to standard error. You can customize the log level by turning on the -vb (verbose) option, so that all log entries will be output to standard error as well as the log file. For example:

python interface.py -vb identify --pathfile="./music/snippet/Track54.wav" --type=2

Example

cd freezam
# create music database
python interface.py -vb construct
# identify a snippet (pre-downloaded)
python interface.py -vb identify --pathfile="./music/snippet/Track54.wav" --type=2

Running the tests

To run the automated tests for this application:

cd shazam
pytest -v test_shazam.py