NCJ Number
195101
Date Published
2002
Length
249 pages
Annotation
This book addresses the basic sciences underlying voice identification, techniques and methods for analyzing recorded evidence, the latest relevant technology, skills such as how to coordinate "ear witness" line-ups, and the appropriate time to call in a professional consultant.
Abstract
One chapter reviews the history of speaker identification (SPID), concluding with a description of "voiceprints," which came on the American scene in the 1960's. The author notes that the harm "voiceprints" brought to law enforcement and the courts and the misuses of the procedure are now obvious. The emphasis has shifted to speaker verification (SV). A chapter on aural-perceptual approaches focuses on the listener, as the speaker is of consequence only when his/her behavior affects that of the listener. Issues discussed include a listener's ability to accurately remember a voice over varying periods of time, the strengths and weaknesses exhibited by the human auditory system, and elements (acoustic and otherwise) related to the environment and the nature of the speaker. The next chapter includes descriptions of the professionals who conduct speaker identification evaluations. A chapter on "earwitness line-ups" notes that problems associated with such a procedure are of growing concern. This form of SPID is becoming more common than in the past, and in some cases the procedure is being conducted by individuals who are only marginally competent. Adequate criteria for their proper use have not as yet been established. In a chapter on voiceprints, the author discusses the merits of Majewski's and Basztura's approach to forensic SPID. They combine aural-perceptual assessment with several pattern-matching spectrographic techniques and multiple computer-based algorithms. The author believes this procedure is useful for law enforcement, at least until more sophisticated approaches are developed. The author's own semi-automatic speaker identification system (SAUSI) is referenced throughout the book, and a full description and discussion of it are provided in the final chapter. SAUSI currently consists of four vectors: long-term spectra, speaking fundamental frequency, time-energy distribution, and vowel formant tracking. Assessment of SAUSI has shown it to be a useful, if not a definitive, system. It has exceeded the "80-percent correct" identification rule-of-thumb minimum and has done so for years. It uses an internal validation procedure, and its application procedures permit stable results to be obtained. Chapter references, suggested readings, and author and subject indexes