Likelihood ratio-based forensic semi-automatic speaker identification with alveolar fricative spectra in a real-world case

Phil Rose

Wednesday, December 14th, 2022, Special Session 11am – 12.30pm


 A real-world forensic voice identification is described in a case involving the blowing-up of a car, three suspects, and a miniscule amount of speech evidence. Necessary stages in the estimation of a likelihood ratio are described, based on the bandlimited cepstral spectral acoustics of two alveolar fricatives /s/ and /z/ in the questioned utterance. The speech evidence is shown to be very much more likely assuming one of the suspects said the utterance.