Elias Pampalk - MA Toolbox


27 Apr 2007
I've added the G1C implementation which I described in my thesis and which did well in the MIREX'06 audio similarity evaluation. More comments on this can found here.

09 Apr 2007
Thanks to helpful comments from Charles Walker and Abhinav Singh I fixed the broken link to emd.zip and updated ma_simple_eval (adding a little more feedback and error checks). There might be an issue with ma_fc (if you get a strange error pointing to line 114 in ma_fc that might be it). Unfortunately, I can't reproduce it and I don't have time to look into it right now. I'm guessing that it happens when there is only one item assigned to a cluster (something Mark Godfrey noticed recently). If that happens you could try using longer audio files (e.g. at least 60 seconds) and fewer clusters.

19 Mar 2007
MA toolbox website moved here.
A bug reported by Mark T. Godfrey in ma_fc was fixed. The bug effected kmeans clustering. Instead of selecting the best cluster model from 5 different runs each with a different random initializations, only one run was used.

22 Jun 2006
Thanks to Adam Berenzweig (ma_cms) and Klaus Seyerlehner (ma_mfcc & ma_fp) a number of bugs were fixed. The one in ma_cms was a major one which effected the KL computations.


The Music Analysis (MA) toolbox is a collection of functions for Matlab. It contains functions to analyze music (audio) and compute similarities. The type of similarity computed can be used to, e.g., generate playlists or organize music collections. For details see the documentation.

The MA Toolbox can be redistributed and/or modified under the terms of the GNU General Public License as published by the Free Software Foundation. Either version 2, or any later version apply.

For full functionality the MA toolbox requires the NETLAB toolbox and the EMD functions by Y. Rubner to be installed. (See download section for details.)

This toolbox will be maintained on a very irregular basis. Any contributions are very welcome!

Related Publications

A Music Similarity Function Based on Signal Analysis
B. Logan and A. Salomon. In the Proceedings of the IEEE International Conference on Multimedia and Expo (ICME'01), Tokyo, August 22-25, 2001.

Music Similarity Measures: What's the Use?
J.-J. Aucouturier and F. Pachet. In the Proceedings of the ISMIR International Conference on Music Information Retrieval (ISMIR'02), Paris, October 13-17, 2002.

Content-based Organization and Visualization of Music Archives
E. Pampalk, A. Rauber, and D. Merkl, In the Proceedings of the ACM Multimedia 2002, Juan les Pins, France, December 1-6, 2002, ACM. PDF.

Exploring Music Collections by Browsing Different Views
E. Pampalk, S. Dixon, and G. Widmer. Proceedings of the ISMIR International Conference on Music Information Retrieval (ISMIR'03), pp 201-208, Baltimore, MD, October 26-30, 2003.

Improving Timbre Similarity: How high is the sky?
J.-J. Aucouturier and F. Pachet. Journal of Negative Results in Speech and Audio Sciences, 1(1), 2004.

A Matlab Toolbox to Compute Similarity from Audio
E. Pampalk. Proceedings of the ISMIR International Conference on Music Information Retrieval (ISMIR'04), Barcelona, Spain, 2004. PDF.

Computational Models of Music Similarity and their Application to Music Information Retrieval
E. Pampalk. Doctoral Thesis, Vienna University of Technology, Austria, March 2006. PDF.

See my list of publications for more publications on computational models of music similarity.

Last update: 27 May 2007