Musher

Simple musical key detection.

Table of Contents

• Musher
• Table of Contents
• Quick Start
  • Install
  • Documentation
  • Sample Usage
  • What key profile should I use?
• Installation
  • Python 3.5+
  • C++ 14
    • Dependencies
      • Cmake
      • Conan
    • Install (WIP)
• Usage
  • Key Profiles
• Development
  • Python
    • Install normally
    • Install in debug mode
  • C++
    • Build normally
    • Build in debug mode
• Tests
  • Python
    • Required dependencies
    • Running tests
  • C++
    • Required dependencies
    • Running tests
• Documentation
  • Link to documentation
  • Dependencies
  • How to Generate
  • Publish to Github Pages
• Cleanup
• Useful links
• Credits

Quick Start

Install

pip install musher 

Documentation

Sample Usage

What key profile should I use?

Installation

You can choose to install this package in 2 ways, Python and C++. The package was primary written in C++ and wrapped in Python. Choose either of the following methods of installation.

Python 3.5+

pip install musher

C++ 14

Dependencies

Cmake

-- MacOS --
brew install cmake

-- Linux --
sudo apt install cmake

-- Windows --
Download from https://cmake.org/download/

Conan

pip install conan

Install (WIP)

conan install musher

Usage

import os
import musher

audio_file_path = os.path.join("data", "audio_files", "mozart_c_major_30sec.wav")
wav_decoded = musher.decode_wav_from_file(audio_file_path)
print(wav_decoded)
{
  'avg_bitrate_kbps': 1411,
  'bit_depth': 16,
  'channels': 2,
  'file_type': 'wav',
  'length_in_seconds': 30.0,
  'mono': False,
  'normalized_samples': array([
        [ 0.        ,  0.        ,  0.        , ..., -0.33203125, -0.32833862, -0.3274536 ],
        [ 0.        ,  0.        ,  0.        , ..., -0.29162598, -0.27130127, -0.25457764]
      ], dtype=float32),
  'sample_rate': 44100,
  'samples_per_channel': 1323000,
  'stereo': True
}

normalized_samples = wav_decoded["normalized_samples"]
sample_rate = wav_decoded["sample_rate"]
# Different key profiles produce different results.
# 'Temperley' produces good results for classical music.
key_profile_type = "Temperley"
key_output = musher.detect_key(normalized_samples, sample_rate, key_profile_type)
print(key_output)
{
  'first_to_second_relative_strength': 0.6078072169442225,
  'key': 'C',
  'scale': 'major',
  'strength': 0.7603224296919142
}

Key Profiles

Description of all available key profiles

Development

Python

Install normally

pip install .
# OR
python3 setup.py install

Install in debug mode

pip install -e .
# OR
python3 setup.py develop

NOTE: This package has 2 dependencies: Pybind11 and Numpy. You may need to install these python packages if the setup.py does not do it for you.

C++

Build normally

python setup.py cmake

# OR

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
cmake --build .

Build in debug mode

python setup.py cmake --debug

# OR

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Debug -DENABLE_TESTS=On
cmake --build .

Tests

Python

Required dependencies

pip install tox pytest

Running tests

This will run tests against a pip installed copy of the source code (best to do before deploying code).

# Running this command does not require pytest to be installed
tox 

This will run tests directly (best to do while modifying the code base).

pytest ./tests -v

C++

Required dependencies

1. Google Test (Conan should install this for you when the project is built in debug mode.)

Running tests

# Ctest
python setup.py ctest

# OR

# Google test
python setup.py gtest

Documentation

Generate documentation using Doxygen, Breathe, and Sphinx.

Link to documentation

Dependencies

1. Doxygen >1.5.1 - Generates C++ documentation
2. Breathe - Translate Doxygen docs to Sphinx docs
3. Sphinx - Generates python documentation
4. C++ dependencies - Required to build the project

How to Generate

python setup.py cmake --docs

Publish to Github Pages

You must install the python module and generate the docs BEFORE running this command.

1. pip install -e .
2. Generate docs

python setup.py publish_docs -m "Added docs for new function"

Cleanup

This will remove all the temporary files/folders created from building and testing this package.

python setup.py clean

Useful links

Python audio libraries:

• https://github.com/vinta/awesome-python#audio

Credits

Essentia - C++ library for audio and music analysis, description and synthesis, including Python bindings.