The Northwestern University Source Separation Library (nussl) (pronounced "nuzzle") is a flexible, object oriented python audio source separation library created by the Interactive Audio Lab at Northwestern University. At its core, nussl provides implementations of common source separation algorithms as well as an easy-to-use framework for prototyping and adding new algorithms. The aim of nussl is to create a low barrier to entry for using popular source separation algorithms, while also allowing the user fine tuned control of low-level parameters.
NOTICE: nussl is currently in alpha. Please be mindful.
Please see the issues page before contacting the authors.
- Master contains the most recent stable version, the version that you get when you pip install nussl.
- Development contains the most recent work, but is not as stable as master. Most all of development work happens in this branch before being pushed to master.
- Experimental has many more implementations, but many have not been vetted or properly tested. Some methods in this branch require extra functionality that cannot be included in a pip install, such as the Vamp binary files or tensorflow. This branch is the bleeding edge.
- gh-pages is used to auto generate our documentation using Sphinx.
- Other feature branches do exist from time to time as well.
Full documentation is available here.
Note: This package has been tested with python 2.7, but not python 3.x yet. Use python 3 at your own peril!
At its core, nussl contains implementations of the following source separation algorithms:
Spatialization algorithms:
- Degenerate Unmixing Estimation Technique (DUET)
- PROJET
Median filtering algorithms:
- REpeating Pattern Extraction Technique (REPET)
- REPET using the cosine similarity matrix (REPET-SIM)
- Harmonic/Percussive Source Separation (HPSS)
- Kernel Adaptive Modeling (KAM)
General matrix decomposition:
- Non-negative Matrix Factorization (NMF)
- Robust Principal Component Analysis (RPCA)
- Independent Component Analysis (ICA)
Other Separation Methods
- Ideal Mask
- Overlap Add
- Algorithm Picker (multicue separation)
- Separation via 2DFT
- Melody Tracking separation (Melodia)
Also coming: Deep methods (Deep Clustering, Huang et. al. Deep Separation, etc.) (with keras backend), Separation by Segmentation, Melody tracking methods (Melodia, SIMM), and many more! Note: newly algorithms usually do not live in the master branch.
The nussl also has an evaluation framework, which provides interoperability between nussl and mir_eval (a python implementation of BSSEval) as well as implementations of other evaluation methods.
We recommend getting installing anaconda prior to using this package.
Installation is easy if you have pip (recommended):
$ pip install nussl
A note on cloning from github: nussl contains a large number of uncompressed audio benchmark files for testing which makes cloning directly from Github a slow process. nussl is now synced with Git LFS, so if you want to clone from Github without downloading the large testing files, please use Git LFS.
Camera ready (in master and pip builds):
- Repet
- RepetSim
- IdealMask
- OverlapAdd
- DUET
In Development Branch:
- NMF
- RPCA
- PROJET
- 2DFT modulation separation
Planned:
- KAM
- HPSS
- SIMM
- Separation via segmentation
- Deep clustering
- Your algorithm? (contact us for details)
nussl 0.1.5* is under an MIT License
The MIT License (MIT)
Copyright (c) 2017 Interactive Audio Lab
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Ethan Manilow (website), Prem Seetharaman (website), Fatemah Pishdadian (website)
Former:
Corey Grief (website),
Contact Ethan Manilow (ethanmanilow [at] u.northwestern.edu) with any questions or issues. Please look at the "issues" page before reporting problems.