Forensic voice comparison and the paradigm shift

Abstract

We are in the midst of a paradigm shift in the forensic comparison sciences. The new paradigm can be characterised as quantitative data-based implementation of the likelihood-ratio framework with quantitative evaluation of the reliability of results. The new paradigm was widely adopted for DNA profile comparison in the 1990s, and is gradually spreading to other branches of forensic science, including forensic voice comparison. The present paper first describes the new paradigm, then describes the history of its adoption for forensic voice comparison over approximately the last decade. The paradigm shift is incomplete and those working in the new paradigm still represent a minority within the forensic-voice-comparison community.

Impact of Background Noise in Mobile Phone Networks on Forensic Voice Comparison

“…The use of the LR for evaluating the speech forensic evidence is increasingly gaining acceptance among forensic scientists [18][19][20][21][22]. The LR framework provides a quantitative answer to the question: How much more likely is it to observe the properties of the offender and suspect speech samples assuming they have the same origin (prosecution hypothesis) than a different origin (defence hypothesis).…”

“…It penalizes the experimental results of same-and different-speaker comparisons that deviate from the actual output and thus provides a measure of accuracy for the system used. Mathematically, C llr can be expressed as [20][21][22][23][24]. Low Cllr values indicate that a FVC analysis system is providing some useful information (the lower the value, the more accurate the analysis, and vice versa).…”

An empirical estimate of the precision of likelihood ratios from a forensic-voice-comparison system

“…A likelihood ratio less than one lends support to the different-speaker hypothesis, e.g., if the trier of fact is provided with a likelihood ratio of 1/100 based on the voice evidence, then, whatever their belief before the evidence was presented, they should now be one hundred times more likely than before to believe that the two voice samples were produced by different speakers. The likelihood ratio framework is recommended by numerous forensic statisticians and forensic scientists, both for forensic comparison in general [1][2][3][4][5][6][7] and for forensic voice comparison in particular [8][9][10][11][12][13][14][15]. Presently, the issue of validity and reliability is of great concern in forensic science [16][17][18][19].…”

An investigation of supervector regression for forensic voice comparison on small data

“…The likelihood ratio is defined as the likelihood that the evidence would be observed if the same-origin hypothesis were true divided by the likelihood that the evidence would be observed if the different-origin hypothesis were true [7][8][9][10][11].…”