F-ratio Client-Dependent Normalisation for Biometric Authentication Tasks. Poh, N. & Bengio, S. In IEEE International Conference on Acoustic, Speech, and Signal Processing, ICASSP, pages 721–724, 2005. ![F-ratio Client-Dependent Normalisation for Biometric Authentication Tasks [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper abstract bibtex This study investigates a new \emphclient-dependent normalisation to improve biometric authentication systems. There exists many client-de-pendent score normalisation techniques applied to speaker authentication, such as Z-Norm, D-Norm and T-Norm. Such normalisation is intended to adjust the variation across different client models. We propose ``F-ratio'' normalisation, or F-Norm, applied to face and speaker authentication systems. This normalisation requires only that \emphas few as two client-dependent accesses are available (the more the better). Different from previous normalisation techniques, F-Norm considers the client and impostor distributions \emphsimultaneously. We show that F-ratio is a natural choice because it is directly associated to Equal Error Rate. It has the effect of centering the client and impostor distributions such that a global threshold can be easily found. Another difference is that F-Norm actually ``interpolates'' between client-independent and client-dependent information by introducing a mixture parameter. This parameter \emphcan be optimised to maximise the class dispersion (the degree of separability between client and impostor distributions) while the aforementioned normalisation techniques cannot. The results of 13 unimodal experiments carried out on the XM2VTS multimodal database show that such normalisation is advantageous over Z-Norm, client-dependent threshold normalisation or no normalisation.
@inproceedings{poh:2005:icassp,
author = {N. Poh and S. Bengio},
title = {F-ratio Client-Dependent Normalisation for Biometric Authentication Tasks},
booktitle = {{IEEE} International Conference on Acoustic, Speech, and Signal Processing, {ICASSP}},
year = 2005,
pages = {721--724},
url = {publications/ps/poh_2005_icassp.ps.gz},
pdf = {publications/pdf/poh_2005_icassp.pdf},
djvu = {publications/djvu/poh_2005_icassp.djvu},
idiap = {publications/pdf/rr04-46.pdf},
original = {2005/fratio_icassp},
topics = {multimodal, biometric_authentication},
web = {http://dx.doi.org/10.1109/ICASSP.2005.1415215},
abstract = {This study investigates a new \emph{client-dependent normalisation} to improve biometric authentication systems. There exists many client-de-pendent score normalisation techniques applied to speaker authentication, such as Z-Norm, D-Norm and T-Norm. Such normalisation is intended to adjust the variation across different client models. We propose ``F-ratio'' normalisation, or F-Norm, applied to face and speaker authentication systems. This normalisation requires only that \emph{as few as} two client-dependent accesses are available (the more the better). Different from previous normalisation techniques, F-Norm considers the client and impostor distributions \emph{simultaneously}. We show that F-ratio is a natural choice because it is directly associated to Equal Error Rate. It has the effect of centering the client and impostor distributions such that a global threshold can be easily found. Another difference is that F-Norm actually ``interpolates'' between client-independent and client-dependent information by introducing a mixture parameter. This parameter \emph{can be optimised} to maximise the class dispersion (the degree of separability between client and impostor distributions) while the aforementioned normalisation techniques cannot. The results of 13 unimodal experiments carried out on the XM2VTS multimodal database show that such normalisation is advantageous over Z-Norm, client-dependent threshold normalisation or no normalisation.},
categorie = {C},
}
Downloads: 0
{"_id":"Z9Sx3PZ36gNb4ThRr","bibbaseid":"poh-bengio-fratioclientdependentnormalisationforbiometricauthenticationtasks-2005","authorIDs":[],"author_short":["Poh, N.","Bengio, S."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["N."],"propositions":[],"lastnames":["Poh"],"suffixes":[]},{"firstnames":["S."],"propositions":[],"lastnames":["Bengio"],"suffixes":[]}],"title":"F-ratio Client-Dependent Normalisation for Biometric Authentication Tasks","booktitle":"IEEE International Conference on Acoustic, Speech, and Signal Processing, ICASSP","year":"2005","pages":"721–724","url":"publications/ps/poh_2005_icassp.ps.gz","pdf":"publications/pdf/poh_2005_icassp.pdf","djvu":"publications/djvu/poh_2005_icassp.djvu","idiap":"publications/pdf/rr04-46.pdf","original":"2005/fratio_icassp","topics":"multimodal, biometric_authentication","web":"http://dx.doi.org/10.1109/ICASSP.2005.1415215","abstract":"This study investigates a new \\emphclient-dependent normalisation to improve biometric authentication systems. There exists many client-de-pendent score normalisation techniques applied to speaker authentication, such as Z-Norm, D-Norm and T-Norm. Such normalisation is intended to adjust the variation across different client models. We propose ``F-ratio'' normalisation, or F-Norm, applied to face and speaker authentication systems. This normalisation requires only that \\emphas few as two client-dependent accesses are available (the more the better). Different from previous normalisation techniques, F-Norm considers the client and impostor distributions \\emphsimultaneously. We show that F-ratio is a natural choice because it is directly associated to Equal Error Rate. It has the effect of centering the client and impostor distributions such that a global threshold can be easily found. Another difference is that F-Norm actually ``interpolates'' between client-independent and client-dependent information by introducing a mixture parameter. This parameter \\emphcan be optimised to maximise the class dispersion (the degree of separability between client and impostor distributions) while the aforementioned normalisation techniques cannot. The results of 13 unimodal experiments carried out on the XM2VTS multimodal database show that such normalisation is advantageous over Z-Norm, client-dependent threshold normalisation or no normalisation.","categorie":"C","bibtex":"@inproceedings{poh:2005:icassp,\n author = {N. Poh and S. Bengio},\n title = {F-ratio Client-Dependent Normalisation for Biometric Authentication Tasks},\n booktitle = {{IEEE} International Conference on Acoustic, Speech, and Signal Processing, {ICASSP}},\n year = 2005,\n pages = {721--724},\n url = {publications/ps/poh_2005_icassp.ps.gz},\n pdf = {publications/pdf/poh_2005_icassp.pdf},\n djvu = {publications/djvu/poh_2005_icassp.djvu},\n idiap = {publications/pdf/rr04-46.pdf},\n original = {2005/fratio_icassp},\n topics = {multimodal, biometric_authentication},\n web = {http://dx.doi.org/10.1109/ICASSP.2005.1415215},\n abstract = {This study investigates a new \\emph{client-dependent normalisation} to improve biometric authentication systems. There exists many client-de-pendent score normalisation techniques applied to speaker authentication, such as Z-Norm, D-Norm and T-Norm. Such normalisation is intended to adjust the variation across different client models. We propose ``F-ratio'' normalisation, or F-Norm, applied to face and speaker authentication systems. This normalisation requires only that \\emph{as few as} two client-dependent accesses are available (the more the better). Different from previous normalisation techniques, F-Norm considers the client and impostor distributions \\emph{simultaneously}. We show that F-ratio is a natural choice because it is directly associated to Equal Error Rate. It has the effect of centering the client and impostor distributions such that a global threshold can be easily found. Another difference is that F-Norm actually ``interpolates'' between client-independent and client-dependent information by introducing a mixture parameter. This parameter \\emph{can be optimised} to maximise the class dispersion (the degree of separability between client and impostor distributions) while the aforementioned normalisation techniques cannot. The results of 13 unimodal experiments carried out on the XM2VTS multimodal database show that such normalisation is advantageous over Z-Norm, client-dependent threshold normalisation or no normalisation.},\n categorie = {C},\n}\n\n","author_short":["Poh, N.","Bengio, S."],"key":"poh:2005:icassp","id":"poh:2005:icassp","bibbaseid":"poh-bengio-fratioclientdependentnormalisationforbiometricauthenticationtasks-2005","role":"author","urls":{"Paper":"http://bengio.abracadoudou.com/publications/ps/poh_2005_icassp.ps.gz"},"downloads":0},"bibtype":"inproceedings","biburl":"http://bengio.abracadoudou.com/samy.bib","creationDate":"2020-03-18T03:43:27.398Z","downloads":0,"keywords":[],"search_terms":["ratio","client","dependent","normalisation","biometric","authentication","tasks","poh","bengio"],"title":"F-ratio Client-Dependent Normalisation for Biometric Authentication Tasks","year":2005,"dataSources":["9NCW2CDr4M3s5DvNX"]}