Meter-2000 dataset

This dataset consists of 2000 song fragments which come with a meter annotation. The meters are 2, 3, 4 or 6. The dataset has been used in the research on meter classification on the basis of audio by Matthias Varewyck.


Meter analysis data

Below you find the audio fragments and multi-level annotations used for the meter analysis research. You can first view the readme file providing a content description or immediately download the zip file (382MByte) that contains the audio and annotation files.

Terms of use

You can use the meter annotation files for research purposes or personal use only. Commercial exploitation of any kind is not allowed. If you would like to make the data available to other interested people, you should point them to this web site and they should read and agree to these terms of use in their own turn. In publications where you make use of this data, you should mention the provider (Ghent University) and refer to the paper "Assessment of state-of-the-art meter analysis algorithms with an extended meter description model, M. Varewyck and J.-P. Martens in ISMIR proceedings 2007". The annotator tried very hard to get the annotations as accurate and consistent as possible, but still the data is provided as-is, without warranty. Downloading these files implies that you have read this text and that you agree to the terms of use.


COGEN speech corpus

The COGEN speech corpus contains spoken dutch material from 174 persons living in Flanders, the Dutch speaking part of Belgium. The corpus was recorded and annotated in a collaborative project with the Speech Lab of ESAT-KULeuven.