The problem of F0 and real-life speaker identification: a case study

Abstract

While it is well known that fundamental frequency of voice can be an important cue in speaker identification and verification in the laboratory, its use is not so effective in the field. This idea is supported by the report of a relevant case. Specifically, the holdup of a gas station by an adult male was video taped by a monitoring camera. When the voice of the criminal was measured, it was found to be very high-pitched; his mean fundamental frequency was 228 Hz. Later, when the criminal was arrested, his 'normal' mean fundamental frequency was determined to be only 140 Hz. The difference here is of considerable magnitude. It is a relationship which often is found in the forensic field. That is, we have come to expect differences between the original and the reference samples of a speaker's mean fundamental frequency because, when the original material is recorded, the speaker can be under highly stressed emotional states and this occurrence has a great influence on his or her F0. These possible situational differences must be taken into account in forensic investigations.