Vowel inherent spectral change in forensic voice comparison.

Abstract

Two-parameter models of vowel inherent spectral change, such as dual-target or target-plus-slope models, have been found to be adequate for vowel-phoneme identification. More sophisticated curve-fitting models do not appear to outperform such two-parameter models. This suggests that if only simple cues such as initial and final formant values are necessary for signaling phoneme identity, then speakers may have considerable freedom in the path taken between the initial and final formant values. If the constraints on formant trajectories are relatively lax with respect to vowel-phoneme identity, then with respect to speaker identity there may be considerable information contained in the details of formant trajectories. Differences in physiology and idiosyncracies in the use of motor commands may mean that different individuals consistently produce different formant trajectories between the beginning and end of the same vowel phoneme. If within-speaker variance is substantially smaller than between-speaker variance, then formant trajectories may be exploited for forensic voice comparison. This paper reviews a number of forensic-voice-comparison studies (including studies conducted using the likelihood-ratio framework) which have extracted information relevant to speaker identity from formant trajectories. For the purposes of forensic voice comparison, models using parametric curves are found to outperform simple two-parameter models.