The corpus described here is a sub set of the Magnatagatune audio database. The latter can be found at the following URL :
It basically consists in a collection of 25860 audio excerpts of 29.3 seconds, encoded in MP3 at 32Kbps bitrate, sampled at 16 Hz, mono. The audio samples are under a Creative Commons Attribution - Noncommercial-Share Alike 3.0 Unported License, and come along with a large amount of human-generated tags and annotations.
We only use a subset of 500 audio samples to perform distortion measures on the fingerprint codes; these samples are randomly drawn in the dataset, and are listed in the following file: