Forensic Speaker Comparison Using Evidence Interval in Full Bayesian Significance Test

This paper describes the application of a full Bayesian significance test (FBST) to compute evidence intervals in forensic speaker comparison (FSC). In the FBST approach, the challenge is to apply the test to a large number of observations and to formulate an equation to solve the test quickly. The contribution of the present work is that it proposes an application of the FBST to FSC and develops a method to calculate the FBST for the distribution of expected values (mean) with unknown variance without using Monte Carlo Markov chains (MCMC). Comparisons with other interval inference methodologies indicate that the evidence interval size is 49% greater than that computed with the Gosset approach. The evidence interval presented 71% fewer classification errors than the punctual inference did for the signal-to-noise ratio (SNR) of 17 dB.

[1]  William M. Bolstad,et al.  Introduction to Bayesian Statistics , 2004 .

[2]  H. Jeffreys.  An invariant form for the prior probability in estimation problems , 1946, Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences.

[3]  Geoffrey Stewart Morrison,et al.  Measuring the validity and reliability of forensic likelihood-ratio systems. , 2011, Science & justice : journal of the Forensic Science Society.

[4]  Natalia L. Oliveira,et al.  A discussion on significance indices for contingency tables under small sample sizes , 2016, PloS one.

[5]  Ruili Wang,et al.  Speaker identification features extraction methods: A systematic review , 2017, Expert Syst. Appl..

[6]  R Togneri,et al.  An Overview of Speaker Identification: Accuracy and Robustness Issues , 2011, IEEE Circuits and Systems Magazine.

[7]  Carlos Alberto de Bragança Pereira,et al.  Can a Significance Test Be Genuinely Bayesian , 2008 .

[8]  Philip Rose,et al.  An empirical estimate of the precision of likelihood ratios from a forensic-voice-comparison system. , 2011, Forensic science international.

[9]  Gaston H. Gonnet,et al.  On the LambertW function , 1996, Adv. Comput. Math..

[10]  A. Silva,et al.  Corpus CEFALA-1: Base de dados audiovisual de locutores para estudos de biometria, fonética e fonologia / Corpus CEFALA-1: Audiovisual Database of Speakers for Biometric, Phonetic and Phonology Studies , 2019, REVISTA DE ESTUDOS DA LINGUAGEM.

[11]  Douglas A. Reynolds,et al.  Robust text-independent speaker identification using Gaussian mixture speaker models , 1995, IEEE Trans. Speech Audio Process..

[12]  John H. L. Hansen,et al.  Speaker Recognition by Machines and Humans: A tutorial review , 2015, IEEE Signal Processing Magazine.

[13]  Julio Michael Stern,et al.  Evidence and Credibility: Full Bayesian Significance Test for Precise Hypotheses , 1999, Entropy.

[14]  Geoffrey Stewart Morrison,et al.  Forensic voice comparison and the paradigm shift. , 2009, Science & justice : journal of the Forensic Science Society.

[15]  Wonyong Sung,et al.  A statistical model-based voice activity detection , 1999, IEEE Signal Processing Letters.

[16]  Cachimo Combo Assane,et al.  Model choice in separate families: A comparison between the FBST and the Cox test , 2017, Commun. Stat. Simul. Comput..

[17]  Julio Michael Stern,et al.  Bayesian evidence test for precise hypotheses , 2003 .

[18]  Peter French,et al.  International practices in forensic speaker comparison , 2011 .

[19]  Jonathan J. Koehler,et al.  The Individualization Fallacy in Forensic Science Evidence , 2008 .

[20]  Geoffrey Stewart Morrison,et al.  A comparison of procedures for the calculation of forensic likelihood ratios from acoustic-phonetic data: Multivariate kernel density (MVKD) versus Gaussian mixture model-universal background model (GMM-UBM) , 2011, Speech Commun..

[21]  S. Zabell,et al.  On Student's 1908 Article “The Probable Error of a Mean” , 2008 .