Music structure segmentation based on shift-invariant probabilistic latent component analysis of chroma
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


SIPLCA Music Analysis

This package contains the implementation of the algorithm for
identifying repeated harmonic patterns in music and structure
segmentation described in [1].  It is primarily written in Python but
calls Matlab for feature extraction and performance evaluation.

It depends on NumPy and SciPy (, Matplotlib
( for plotting, and the mlabwrap
package ( for running Matlab code.

Included is Dan Ellis' beat-synchronous chroma feature extraction
code, also available from:

Also useful (but not included) is Dan's code for resynthesizing

All code in this package is distributed under the terms of the GNU
General Public License.  See the file COPYING for details.

Example usage

From the command line
````````````````````` can be run as a script to run the structure segmentation
algorithm on a single audio file.  The usage is as follows: -i inputfile.wav -o outputfile [-param1 val1] [-param2 val2] ...

Where parameters "-param1", etc. are optional and are defined in the
docstring for segmenter.segment_song.
For MIREX 2010, please run the shell script as follows: inputfile.wav outputfile

From Python
>>> import segmenter
>>> wavfile = '/path/to/come_together.wav'
>>> rank = 4  # rank corresponds to the number of segments
>>> win = 60  # win controls the length of each chroma pattern
>>> niter = 200  # number of iterations to perform
>>> np.random.seed(123)  # Make this reproduceable
>>> labels = segmenter.segment_wavfile(wavfile, win=win, rank=rank,
...                                    niter=niter, plotiter=10)
INFO:plca:Iteration 0: divergence = 10.065992
INFO:plca:Iteration 50: divergence = 9.468196
INFO:plca:Iteration 100: divergence = 9.421632
INFO:plca:Iteration 150: divergence = 9.409279
INFO:root:Iteration 199: final divergence = 9.404961
INFO:segmenter:Removing 2 segments shorter than 32 frames

.. image::come_together-segmentation.png

>>> print labels
0.0000 21.7480 segment0
21.7480 37.7640 segment1
37.7640 55.1000 segment0
55.1000 76.1440 segment1
76.1440 95.1640 segment0
95.1640 121.2360 segment1
121.2360 158.5360 segment2
158.5360 180.8520 segment1
180.8520 196.5840 segment0
196.5840 255.8160 segment3

See the docstrings in and for more detailed

This material is based upon work supported by the NSF (grant
IIS-0844654) and by the IMLS (grant LG-06-08-0073-08).


 [1] R. J. Weiss and J. P. Bello. "Identifying Repeated Patterns in
     Music Using Sparse Convolutive Non-Negative Matrix
     Factorization". In Proc. International Conference on Music
     Information Retrieval (ISMIR), 2010.

2010-05-27 Ron Weiss <>