SHEEP, GOATS, LAMBS and WOLVES. A statistical analysis of speaker performance in the NIST 1998 Speaker Recognition Evaluation. Doddington, G. R; Liggett, W.; Martin, A. F; Przybocki, M.; and Reynolds, D. A In ICSLP 1998. Proceedings of the 5th International Conference on Spoken Language Processing, pages 1351-1354, Sidney Convention Centre, Sidney, Australia, 30 November - 4 December, 1998.
SHEEP, GOATS, LAMBS and WOLVES. A statistical analysis of speaker performance in the NIST 1998 Speaker Recognition Evaluation [pdf]Paper  abstract   bibtex   
Performance variability in speech and speaker recognition systems can be attributed to many factors. One major factor, which is acknowledged but seldom analyzed, is differences in the recognizability of different speakers. In speaker recognition systems such differences are characterized by the use of animal names for different types of speakers, including sheep, goats, lambs and wolves, depending on their behavior with respect to automatic recognition systems. In this paper we propose statistical tests for the existance of these animals and hunt for such animals using results from the 1998 NIST speaker recognition evaluation.
@inproceedings{doddington_sheep_1998,
	Address = {Sidney Convention Centre, Sidney, Australia, 30 November - 4 December, 1998},
	Author = {Doddington, George R and Liggett, Walter and Martin, Alvin F and Przybocki, Mark and Reynolds, Douglas A},
	Booktitle = {ICSLP 1998. Proceedings of the 5th International Conference on Spoken Language Processing},
	Date = {1998},
	Date-Modified = {2016-09-24 18:56:02 +0000},
	Keywords = {interspeaker variation, speech recognition, speech technology},
	Pages = {1351-1354},
	Title = {SHEEP, GOATS, LAMBS and WOLVES. A statistical analysis of speaker performance in the NIST 1998 Speaker Recognition Evaluation},
	Url = {http://www.itl.nist.gov/iad/mig/publications/storage_paper/icslp_98.pdf},
	Abstract = {Performance variability in speech and speaker recognition systems can be attributed to many factors. One major factor, which is acknowledged but seldom analyzed, is differences in the recognizability of different speakers. In speaker recognition systems such differences are characterized by the use of animal names for different types of speakers, including sheep, goats, lambs and wolves, depending on their behavior with respect to automatic recognition systems. In this paper we propose statistical tests for the existance of these animals and hunt for such animals using results from the 1998 NIST speaker recognition evaluation.},
	Bdsk-Url-1 = {http://www.itl.nist.gov/iad/mig/publications/storage_paper/icslp_98.pdf}}
Downloads: 0