Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability

Michael Wagner; Yuko Kinoshita

Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability

Michael Wagner, Yuko Kinoshita

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

This study investigates the effect of background data specifici-ty on likelihood ratio and prior odds, and consequently on the posterior odds outcome. It is motivated by discussions on the correct choice of speaker recognition background, particularly in forensic voice comparison. We performed strictly controlled experiments with the ANDOSL database where background specificity is the sole independent variable. Results show that target and non-target scores are better separated with less spe-cific background, but that in turn priors must be adjusted down. Because the risk of class recognition instead of individ-ual recognition increases with lower background specificity, we suggest that the prior probability in the Bayes formula is factorised into one part that remains in the domain of the trier of fact as is conventional and another part that is related to the specificity of the assumed or agreed background.

Original language	English
Title of host publication	Proceedings of the Sixteenth Australasian International Conference on Speech Science and Technology
Editors	Christopher Carignan and Michael D. Tyler
Place of Publication	Parramatta, Australia
Publisher	The Australasian Speech Science and Technology Association, Inc.
Pages	353-356
Edition	Peer Reviewed
ISBN (Print)	2207-1296
Publication status	Published - 2016
Event	16th Australasian International Conference on Speech Science and Technology - Parramatta, Australia Duration: 6 Dec 2016 → 9 Dec 2016 https://webarchive.nla.gov.au/awa/20180320002425/http://sst2016.westernsydney.edu.au/ https://webarchive.nla.gov.au/awa/20240426142921/https://assta.org/wp-content/uploads/2019/09/SST2016_Proceedings.pdf

Conference

Conference	16th Australasian International Conference on Speech Science and Technology
Abbreviated title	SST2016
Country/Territory	Australia
City	Parramatta
Period	6/12/16 → 9/12/16
Other	The Australasian Speech Science and Technology Association and Western Sydney University are pleased to host the 16th Australasian International Conference on Speech Science and Technology (SST2016). This conference will mark the 30-year anniversary of the first SST conference held in Canberra in 1986.
Internet address	https://webarchive.nla.gov.au/awa/20180320002425/http://sst2016.westernsydney.edu.au/ https://webarchive.nla.gov.au/awa/20240426142921/https://assta.org/wp-content/uploads/2019/09/SST2016_Proceedings.pdf

Cite this

Wagner, Michael ; Kinoshita, Yuko. / Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability. Proceedings of the Sixteenth Australasian International Conference on Speech Science and Technology. editor / Christopher Carignan and Michael D. Tyler. Peer Reviewed. ed. Parramatta, Australia : The Australasian Speech Science and Technology Association, Inc., 2016. pp. 353-356

@inproceedings{e37ef8f826194c7ebcf11e93b30d1eea,

title = "Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability",

abstract = "This study investigates the effect of background data specifici-ty on likelihood ratio and prior odds, and consequently on the posterior odds outcome. It is motivated by discussions on the correct choice of speaker recognition background, particularly in forensic voice comparison. We performed strictly controlled experiments with the ANDOSL database where background specificity is the sole independent variable. Results show that target and non-target scores are better separated with less spe-cific background, but that in turn priors must be adjusted down. Because the risk of class recognition instead of individ-ual recognition increases with lower background specificity, we suggest that the prior probability in the Bayes formula is factorised into one part that remains in the domain of the trier of fact as is conventional and another part that is related to the specificity of the assumed or agreed background.",

author = "Michael Wagner and Yuko Kinoshita",

year = "2016",

language = "English",

isbn = "2207-1296",

pages = "353--356",

editor = "{Christopher Carignan and Michael D. Tyler}",

booktitle = "Proceedings of the Sixteenth Australasian International Conference on Speech Science and Technology",

publisher = "The Australasian Speech Science and Technology Association, Inc.",

edition = "Peer Reviewed",

note = "16th Australasian International Conference on Speech Science and Technology, SST2016 ; Conference date: 06-12-2016 Through 09-12-2016",

url = "https://webarchive.nla.gov.au/awa/20180320002425/http://sst2016.westernsydney.edu.au/, https://webarchive.nla.gov.au/awa/20240426142921/https://assta.org/wp-content/uploads/2019/09/SST2016_Proceedings.pdf",

}

Wagner, M & Kinoshita, Y 2016, Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability. in CCAMDT (ed.), Proceedings of the Sixteenth Australasian International Conference on Speech Science and Technology. Peer Reviewed edn, The Australasian Speech Science and Technology Association, Inc., Parramatta, Australia, pp. 353-356, 16th Australasian International Conference on Speech Science and Technology, Parramatta, New South Wales, Australia, 6/12/16.

Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability. / Wagner, Michael; Kinoshita, Yuko.
Proceedings of the Sixteenth Australasian International Conference on Speech Science and Technology. ed. / Christopher Carignan and Michael D. Tyler. Peer Reviewed. ed. Parramatta, Australia: The Australasian Speech Science and Technology Association, Inc., 2016. p. 353-356.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability

AU - Wagner, Michael

AU - Kinoshita, Yuko

PY - 2016

Y1 - 2016

N2 - This study investigates the effect of background data specifici-ty on likelihood ratio and prior odds, and consequently on the posterior odds outcome. It is motivated by discussions on the correct choice of speaker recognition background, particularly in forensic voice comparison. We performed strictly controlled experiments with the ANDOSL database where background specificity is the sole independent variable. Results show that target and non-target scores are better separated with less spe-cific background, but that in turn priors must be adjusted down. Because the risk of class recognition instead of individ-ual recognition increases with lower background specificity, we suggest that the prior probability in the Bayes formula is factorised into one part that remains in the domain of the trier of fact as is conventional and another part that is related to the specificity of the assumed or agreed background.

AB - This study investigates the effect of background data specifici-ty on likelihood ratio and prior odds, and consequently on the posterior odds outcome. It is motivated by discussions on the correct choice of speaker recognition background, particularly in forensic voice comparison. We performed strictly controlled experiments with the ANDOSL database where background specificity is the sole independent variable. Results show that target and non-target scores are better separated with less spe-cific background, but that in turn priors must be adjusted down. Because the risk of class recognition instead of individ-ual recognition increases with lower background specificity, we suggest that the prior probability in the Bayes formula is factorised into one part that remains in the domain of the trier of fact as is conventional and another part that is related to the specificity of the assumed or agreed background.

M3 - Conference contribution

SN - 2207-1296

SP - 353

EP - 356

BT - Proceedings of the Sixteenth Australasian International Conference on Speech Science and Technology

A2 - null, Christopher Carignan and Michael D. Tyler

PB - The Australasian Speech Science and Technology Association, Inc.

CY - Parramatta, Australia

T2 - 16th Australasian International Conference on Speech Science and Technology

Y2 - 6 December 2016 through 9 December 2016

ER -

Background Specificity in Forensic Voice Comparison and Its Relation to the Bayesian Prior Probability

Abstract

Conference

Fingerprint

Cite this