The PRISM set is a new database definition for evaluation of speaker recognition systems. The database is created using data from NIST SREs from 2005 to 2010, as well as other data sets available through LDC. This database involves types of variability already seen in NIST speaker recognition evaluations (SREs) like language, channel, speech style and vocal effort, and new types not yet available on any standard database like severe noise, and reverberation. Its design is the result of a collaboration of SRI International and Brno University of Technology during the BEST IARPA program. We also welcome contributions, please help us in making this set better. We chose a liberal license for the content of this set (CC BY-SA) so that modifications will be published under the same license.

If you find this ressource useful, you can cite it in your work.

@article{ferrerpromoting,

title={Promoting robustness for speaker modeling in the community: the PRISM evaluation set}, author={Ferrer, L. and Bratt, H. and Burget, L. and Cernocky, H. and Glembek, O. and Graciarena, M. and Lawson, A. and Lei, Y. and Matejka, P. and Plchot, O. and others} }

Links for the noise samples are hosted by BUT at: http://speech.fit.vutbr.cz/files/PRISM-SET-NOISE.tgz