README

The goal of this corpus is to provide data for music/speech discrimination,
speech/nonspeech detection, and voice activity detection.  The corpus is 
divided into music, speech, and noise portions.  In total there are 
approximately 109 hours of audio. The directories are partitioned by data
source (i.e., the website we downloaded the content from).

Each subdirectory contains a LICENSE file which connects the individual
files in that directory to the governing license as well as attribution
appropriate to the license type. For example, a music entry in a LICENSE 
file may contain the filename, title, artist, a url to the source, and a
summary of the license.  All files in this corpus fall under a Creative 
Commons license or are considered to be in the USA Public Domain.  To 
broaded the use of this corpus, we've avoided including any content which
forbids commercial use. Please refer to http://creativecommons.org/licenses/
for more information about the Creative Commons licenses. 

Most directories contain an ANNOTATIONS file which provide some useful 
metadata. For example, music is annotated for the precense or absence of
vocals and by genre(s). The READMEs in each subdirectory describe the 
annotations in more detail.

Please acknowledge this work if it contributes significantly to any 
publication:

@misc{1510.08484,
  author = {David Snyder and Guoguo Chen and Daniel Povey},
  title = {{MUSAN}: {A} {M}usic, {S}peech, and {N}oise {C}orpus},
  year = {2015},
  eprint = {1510.08484},
  note = {arXiv:1510.08484v1}
}

This work was supported by the National Science Foundation Graduate Research
Fellowship under Grant No. 1232825 and by Spoken Communications. 

David Snyder (email: [email protected])
Guoguo Chen
Daniel Povey