ABSTRACT
This chapter reviews the problems associated with forensic speaker verification and the state of the art of verification algorithms and systems and gives some recent results from the research work done at the BGU in Israel. Computer verification of speakers will have to remain an assist device, coupled with the more conventional methods of trained listeners and manual examination of spectrographs. From forensic point of view the most interesting aspects are those of Emotional State analysis and Speaker Identification and Verification. It is well-known that the human speech conveys not only the intended message but additional voluntary and non-voluntary information concerning the speaker. Speaker verification problems encountered in police work are usually text independent. In speaker verification and identification methods one tries to extract from the given utterance some clues which uniquely relate to the speaker. The chapter concludes that with technology the problems presented by the forensic applications cannot be completely met by automatic machines.
