This presentation provides an introduction to the evaluation and presentation of forensic-comparison evidence using the same framework as is standard for the forensic evaluation of DNA profiles, the likelihood-ratio framework. This framework is recommended as the logically correct framework for the evaluation of all types of evidence involving the comparison of the quantifiable properties of pairs of samples, e.g., DNA profiles, finger marks, glass fragments, and voice recordings. This presentation uses examples from forensic voice comparison.
Given a sample of known origin and a sample of different origin, a forensic likelihood ratio is a quantitative strength-of-evidence statement as to the probability of obtaining the observed differences between two samples under the hypothesis that they have the same origin versus under the hypothesis that they have different origins (e.g., same speaker versus different speaker). A forensic likelihood ratio takes into account both the similarity of the samples and their typicality with respect to the potential population of offenders. Quantitative data-based implementations of the likelihood-ratio framework are objective and their validity and reliability are testable.
The presentation focusses at a conceptual level on the details of the calculation of likelihood ratios in forensic voice comparison.