Department of Software Technology
Vienna University of Technology

The SOMeJB Music Digital Library - Download Page

Many people have asked us to provide them with the implementations we have used for performing our music analysis. So, here we go: below we provide some modules that were used in the context of the SomeJB project for organizing music libraries.
However: Please note, that the present system is a Research Prototype under constant development, extension, modifications, etc. and by no means a production version. You are free to use it for non-commercial purposes (you would not seriously consider using it for commercial purposes anyway) as you please, and we would be happy to learn from your experiences. We will also try to help wherever we can - however, please be aware that we will not be able to provide any guaranteed support for the system :-) If you think you really want to play around with the system in more detail, and possible fix some bugs that we have left there, or plan on extending or adapting it to your needs, we are happy to share the source code and any possible advice we can give with you. In this case, just drop me an e-mail at rauber@ifs.tuwien.ac.at, and I'll do my best to send you whatever you think you want or need.

Available on this page:

Feature Extraction Modules for SOMeJB Prototype 2: Matlab libraries for feature extraction and psycho-acoustic pre-processing.
Feature Extraction Modules for SOMeJB Prototype 1: Modules used together with XMMS to extract features from you rmusic files, as well as a short description of how to get the system running. (discontinued)
Data Clustering and Visualization (SOM/GHSOM/SDH): Links to SOM/GHSOM implementations in C and Matlab as well as Matlab sources for the Smoothed Data Histograms (SDH) used for the experiments and visualization of SOMeJB and Islands of Music (IoM).

Feature Extraction SOMeJB Prototype 2

We provide the Matlab music analysis toolbox implementing the feature extraction and psycho-acoustic pre-processing steps used in the SOM-enhanced Jukebox system.

Matlab Feature Extraction Toolbox:
Matlab music analysis toolbox for feature extraction (rhythm patterns including psycho-acoustic models) from raw audio data. Requires the Matlab signal processing toolbox.
Detailed documentation as well as examples and the most recent version can be found at the
MA Toolbox Project Homepage.
locally cached version (may be slightly out-of-date): MA_MATLAB.tar.gz (gnu-zipped tar archive)

Feature Extraction for SOMeJB Prototype 1

The feature extraction of Prototype 1 is based on the open-source media player xmms. (See the xmms project website for details on obtaining and installing xmms.) A modified visualization plugin is used to extract the frequency data while playing the files. Below, you can either obatin a complete distribution of xmms 1.0.1 together with the anafile plugin, or get the anafile plugin separately and install it into your current xmms path. A (very) short description on how to perform the experiments is provided as well.
More details can be found on the original SOMeJB Prototype 1 project homepage, now discontinued and available only as a locally cached copy in the local projects archive.

XMMS-1.0.1-somejb.tar.gz (xmms + anaflie plugin sources, gnu-zipped tar archive, 732 KB)
anafile_xmms-1.0.1.tar.gz (only the required anafile visualization plugin sources, to be integrated into your xmms installation, gnu-zipped tar archive, 9 KB)
SOMEJB_MODULES.tar.gz (SOMeJB modules sources and handbooks, gnu-zipped tar archive, 474 KB)
somejb_prot1_howto.html (a short HOWTO describing how to set up SOMeJB)

SOM, GHSOM and SDH modules

For organizing the pieces of music on a 2-dimensional map, the Self-Organizing Map (SOM), as well as its hierarchical, fully adaptive extension, the Growing Hierarchical SOM (GHSOM) have been used.

C-Implementation of the GHSOM: The GHSOM is capable of producing of producing (1) "conventional" SOMs, (2) growing SOMs and (The source code provided is tested to compile under Linux, but should work equally well on most other systems. Just do a make configure and then make to compile the system. Otherwise, a compiled binary is provided as well.

The links provided here always link to the latest release of the GHSOM software. For older versions, as well as for more detaile dinformation, property files, etc. refer to the download-page of the GHSOM project
- GHSOM-guide: (HTML-file) A short guide describing the file formats, the parameters of the GHSOM to be specified in the property file, etc.
- GHSOM.tgz: Source code (gnu-zipped tar archive, 210 KB)
- ghsom-binary: (binary LINUX x86)
- Sample property files for:
- somejb2.tv: the template vector file for the 1.200-dimensional input vectors used in the SOMeJB2-Prototype.
- col77_median.in.gz: The input vector file for the Collection 77, described in more detail in the experiments section.
GHSOM MATLAB Toolbox: We also have a Matlab implementation of the GHSOM available, which was developed in a joint project with the University of Aberdeen. Detailed documentation and examples, as well as the most recent version of this toolbox can be found at the GHSOM Matlab Toolbox homepage.
locally cached version (may be slightly out-of-date): GHSOM_MATLAB.tar.gz (gnu-zipped tar archive, 138 KB)
Smoothed Data Histograms (SDH): A Matlab toolbox implementing the smoothed data histograms (SDH) used for creating the Islands of Music visualization in the SOMeJB version 2.
Detailed documentation as well as examples and the most recent version can be found at the SDH Project Homepage.
locally cached version (may be slightly out-of-date): SDH_MATLAB.tar.gz (gnu-zipped tar archive, 59KB)

Up to the SOMeJB Homepage
Comments: rauber@ifs.tuwien.ac.at