Developing Speech Rhythm Analysis for Forensic Voice Comparison

Carroll, Luke and Brown, Georgina (2025) Developing Speech Rhythm Analysis for Forensic Voice Comparison. PhD thesis, Lancaster University.

Text (2025CarrollPhD)
2025CarrollPhD.pdf - Published Version
Download (7MB)

Abstract

Forensic voice comparison (FVC) involves the comparison of a criminal recording (e.g., a threatening phone call), and a known suspect sample (e.g., a police interview). It is the role of an expert forensic analyst to advise the trier of fact (e.g., judge or jury) on the likelihood that the two samples include the same or different speakers. To do this, the expert will carry out an assessment of the similarity of the speech characteristics in the criminal recording and the suspect sample. Speech rhythm has been proposed as a feature that could contribute to FVC, but there is not yet a structured analysis framework that practitioners can exploit in forensic casework. When an analyst suspects a speaker’s speech rhythm is relevant to an analysis, it is usually only described at an impressionistic level. Using both production and perception experiments, the present research explores whether there are acoustic and auditory cues that could capture speech rhythm and subsequently be used to discriminate between speakers in forensic casework. The production experiments revealed that there was very little discriminatory power in syllabic duration, intensity and f₀ measurements across spontaneous, content-mismatched utterances. However, there does appear to be some speaker discriminatory value in applying these same measurements to, so-called, “frequently occurring speech units” (e.g., “er”, “erm”, “yes” and “no”). The perception experiments aimed to determine whether listeners (expert and non-expert) can make meaningful speaker identification assessments when presented with delexicalised speech samples that foreground the rhythmic attributes of speech. Results revealed that expert listeners were better than non-expert listeners in making correct speaker identification assessments, with those who had expertise in forensic phonetics generally performing better than those who did not. The findings from these experiments give promise to the prospect of developing a perceptual (auditory) rhythm framework which can used in forensic casework.