Dear Colleagues, Thank you for your interest in my dissertation: Campbell, J. P., Jr. "Features and Measures for Speaker Recognition." Ph.D. Dissertation, Oklahoma State University, December 1992. Please order a copy of my thesis from UMI: University Microfilms International | UMI Dissertation Services P.O. Box 1346 | 300 North Zeeb Road Ann Arbor, MI 48106-9866 | Ann Arbor, MI 48106-1346 USA | USA (800) 521-0600 | (313) 761-4700 UMI offers hardcover paper (US$ 41.50 academic price), softcover paper, 105mm microfiche, and 35mm rollfilm copies. My intent in publishing my abstract in the Oct '93 IEEE Signal Processing Magazine was not to personally offer copies of my dissertation, but to communicate with other researchers, who I hope will offer constructive criticism as I continue to research speaker verification. I hope this information helps you and I look forward to hearing from you, especially after you have read my dissertation. ----- The divergence (aka symmetric Kullback-Leibler) measure and Bhattacharyya distance are good measures to use between LSP densities; whereas the Euclidean and Mahalanobis distances are mediocre. My modified divergence, called "divergence shape" (which ignores the means and uses only the covariances) between LSP densities outperformed all the other features and measures I tried (e.g., conventional Euclidean distance between LPC cepstrum densities). The Oct '93 IEEE Signal Processing Magazine published an abstract of my work that provides a few more details: Title: Features and Measures for Speaker Recognition Author: Joseph P. Campbell, Jr. Advisor: Rao K. Yarlagadda Granting Institution: Oklahoma State University, Stillwater, OK Acceptance Date: Dec. 1992 Information: Contact the author at the U.S. Government, Department of Defense, Ft. Meade, MD 20755-6000. E-mail: j.campbell@ieee.org This work derives and demonstrates new and powerful features and measures for automatic speaker recognition and compares them with traditional ones. Automatic speaker recognition is the use of a machine to recognize a person from a spoken phrase. Speaker recognition systems can identify a particular person or verify a person's claimed identity. The scope of this study is limited to speech collected from cooperative users in office environments, without adverse microphone or channel impairments. The success of these systems depends directly upon the power of the features and measures used to discriminate between people. The focus of this research is to discover powerful features and measures for speaker verification. After a thorough literature review, concepts were synthesized from such diverse fields as signal processing, information theory, pattern recognition, physiology, and speech production and perception. The most promising innovations were then compared analytically and by computer simulation. New perceptually based features were found which, unfortunately, did not outperform traditional speech production features with respect to speaker identification errors. Powerful new production features and measures for speaker verification were discovered. The main contribution is a new information theoretic measure between the densities of line spectrum pair (LSP) frequency features. This new measure, the divergence shape, can be interpreted geometrically as the shape of an information theoretic measure called divergence. The LSPs were found to be very effective features in this divergence shape measure. On an 87-speaker, closed-set population for a text-independent task, the experimental results show that the LSP densities measured by the divergence shape yield 99.95% speaker identification accuracy. ----- I later discovered (as reported in my ICASSP-95 paper ) that 1.15% error is the identification accuracy of my system when measured in a conventional fashion. Thank you very much for your interest in my work, Joe ____________________________________________________________________________ |Dr. Campbell, US DoD, Johns Hopkins U, Ed. IEEE TSAP, Chair Bio Consortium| | Speaking for myself j.campbell@ieee.org | | Sun Mail Tool attachments |