ESCA Workshop on Automatic Speaker Recognition, Identification, and Verification

Martigny, Switzerland
April 7-9, 1994

Current Approaches to Forensic Speaker Recognition

Hermann J. Künzel

Department of Speaker Identification, Tape Authentication and Linguistic Text Analysis, Bundeskriminalamt, Wiesbaden, Germany

Forensic speaker recognition has become an important tool in crime contravention, since the use of the human voice as an instrument in the commission of crime is ever-increasing. To a certain degree, this is undoubtedly a consequence of the highly-developed and fully automated telephone networks of the industrialized countries, which may safeguard a perpetrator's anonymity almost perfectly. To date, different approaches to forensic speaker identification have been developed by phoneticians, linguists, speech scientists and engineers. In this review, the three main types of procedures currently used under various legal systems will be presented. It is argued that although computer-based tools for acoustical analysis and new statistical data about some speaker-specific features may aid the expert in drawing his conclusions, a fully automatic voice identification device is certainly not in sight, due to the vast number of imponderables and distortions of the speech signal which are typical to the forensic situation.

Full Paper

Bibliographic reference.  Künzel, Hermann J. (1994): "Current approaches to forensic speaker recognition", In ASRIV-1994, 135-142.