I'm trying to find a relatively inexpensive corpus for Speaker
Recognition evaluation, which contains at least 200 talkers (ideally
talking across different conditions or times). The NIST SRE corpora
look promising, but it's unclear to me which one is best, and how many
talkers they contain. If anyone has a good suggestion, please let me
know.
Best Regards,
David
Voxforge is an intriguging possibility, but I'm not sure about how to
download the audio files. All I can find is a browser-based download
of one file at a time. Do you know of a faster way I can download
hundreds of files?
You can use wget. See
http://www.voxforge.org/home/forums/other-languages-forum/general-discussion/corpora-downloading/4
If you have trouble downloading it with wget I can share it as a
single archive