Towards better making a decision in speaker verification

Chen, Ke

doi:10.1016/s0031-3203(02)00034-1

Cited by 52 publications

(21 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Two main approaches can be utilized in this endeavor. The first is known as client-specific decision [9] [16] [3] . This technique substitutes the global match threshold criteria of traditional biometric systems with a decision boundary that is allowed to fluctuate based on individual user match characteristics.…”

Section: Biometric Menagerie Correctionmentioning

confidence: 99%

See 1 more Smart Citation

A statistical investigation into the stability of iris recognition in diverse population sets

Howard

Etter

2014

SPIE Proceedings

View full text Add to dashboard Cite

Iris recognition is increasingly being deployed on population wide scales for important applications such as border security, social service administration, criminal identification and general population management. The error rates for this incredibly accurate form of biometric identification are established using well known, laboratory quality datasets. However, it is has long been acknowledged in biometric theory that not all individuals have the same likelihood of being correctly serviced by a biometric system. Typically, techniques for identifying clients that are likely to experience a false non-match or a false match error are carried out on a per-subject basis. This research makes the novel hypothesis that certain ethnical denominations are more or less likely to experience a biometric error. Through established statistical techniques, we demonstrate this hypothesis to be true and document the notable effect that the ethnicity of the client has on iris similarity scores. Understanding the expected impact of ethnical diversity on iris recognition accuracy is crucial to the future success of this technology as it is deployed in areas where the target population consists of clientele from a range of geographic backgrounds, such as border crossings and immigration check points.

show abstract

Section: Biometric Menagerie Correctionmentioning

confidence: 99%

“…Note that the intra-ethnical distributions are also included for comparison purposes. These subsets were obtained by applying Equation 3 and Equation 2 to the dataset described in Section 4. …”

Section: Inter-ethnical Imposter Comparisonsmentioning

confidence: 99%

A statistical investigation into the stability of iris recognition in diverse population sets

Howard

Etter

2014

SPIE Proceedings

View full text Add to dashboard Cite

show abstract

“…In the literature, the following three ways have been used to deal with the score variability: client-specific threshold, clientspecific fusion and client-specific score normalization [7]. In client-specific threshold, each user has a different threshold [8], which can be a function of a global threshold [9]. However, the decision threshold has to be tuned for each user separately making it difficult for large population applications.…”

Section: λ(X Testmentioning

confidence: 99%

Efficient speaker verification system using speaker model clustering for T and Z normalizations

Ravulakollu

Apsingekar

León

2008

2008 42nd Annual IEEE International Carnahan Conference on Security Technology

View full text Add to dashboard Cite

Abstract-In speaker verification (SV) systems based on Gaussian Mixture Model-Universal Background Model (GMM-UBM), normalization is an important component in the decision stage. Many normalization methods including the T-and Znorms, have been proposed and investigated and these have contributed to state-of-the-art SV systems which have extremely low equal-error rates (EERs). In this paper, we consider application of both T-and Z-norms to a carefully selected subset of speakers using a data driven approach which can significantly reduce computation resulting in faster SV decisions and lower EER. Unfortunately, selection of the subset is critical and must be representative of the entire speaker model space otherwise error rates will increase. In order to properly select the subset of speakers for the normalizations, we propose a novel method which first clusters the speaker models using the K-means algorithm and the Kullback-Leibler (KL) divergence and then selects a set of speakers within the cluster. We evaluate the approach using both the TIMIT, NTIMIT and NIST-2002 corpora and compare against standard T-and Z-normalizations.

show abstract

“…User-specific processing is done either by user-specific normalization [7], [8], user-specific threshold setting [9], user-specific fusion [10], [8] and user-specific template update [11].…”

Section: Introductionmentioning

confidence: 99%

Improving Biometric Verification Systems by Fusing Z-norm and F-norm

Bengherabi

Harizi

Poh

et al. 2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. User-specific score normalization which is related to biometric menagerie has received a lot of attention in the last decade. It is a one-to-one mapping function such that after its application, only a global threshold is needed. In this paper we propose a novel user-specific score normalization framework based on the fusion of Z-norm and F-norm. In this framework, firstly one post-processes the biometric system scores with Z-norm and F-norm procedures. Then, one feeds the resulting two dimensional normalized score vector to a fusion classifier to obtain a final normalized score. Using logistic regression as a fusion classifier, experiments carried out on 13 face and speech systems of the XM2VTS database show that the proposed strategy is likely to improve over the original separate score normalization schemes (F-norm or Z-norm). Furthermore, this novel strategy turns out to be the best strategy for applications requiring low false acceptance rate.

show abstract

Towards better making a decision in speaker verification

Cited by 52 publications

References 27 publications

A statistical investigation into the stability of iris recognition in diverse population sets

A statistical investigation into the stability of iris recognition in diverse population sets

Efficient speaker verification system using speaker model clustering for T and Z normalizations

Improving Biometric Verification Systems by Fusing Z-norm and F-norm

Contact Info

Product

Resources

About