Generative modelling for unsupervised score calibration

Score calibration enables automatic speaker recognizers to make cost-effective accept / reject decisions. Traditional calibration requires supervised data, which is an expensive resource. We propose a 2-component GMM for unsupervised calibration and demonstrate good performance relative to a supervised baseline on NIST SRE'10 and SRE'12. A Bayesian analysis demonstrates that the uncertainty associated with the unsupervised calibration parameter estimates is surprisingly small.

• Publication year

  2013

• Venue

  IEEE International Conference on Acoustics, Speech, and Signal Processing

• Publication date

  2013-11-04

• Fields of study

  Mathematics, Computer Science

• Identifiers
  DOI 10.1109/ICASSP.2014.6853884arXiv 1311.0707
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Pattern Recognition And Machine Learning

  2016cited by this paper
• The distribution of calibrated likelihood-ratios in speaker recognition

  2013influential reference
• Likelihood-ratio calibration using prior-weighted proper scoring rules

  2013cited by this paper
• The BOSARIS Toolkit: Theory, Algorithms and Code for Surviving the New DCF

  2013influential reference
• Analysis of i-vector Length Normalization in Speaker Recognition Systems

  2011cited by this paper
• Resources for new research directions in speaker recognition: the mixer 3, 4 and 5 corpora

  2007cited by this paper
• Fusion of Heterogeneous Speaker Recognition Systems in the STBU Submission for the NIST Speaker Recognition Evaluation 2006

  2007cited by this paper
• Pattern Recognition and Machine Learning

  2006cited by this paper
• Between-Source Modelling for Likelihood Ratio Computation in Forensic Biometric Recognition

  2005cited by this paper
• Application-independent evaluation of speaker detection

  2004cited by this paper
• Information Theory, Inference, and Learning Algorithms

  2004cited by this paper
• Data analysis : a Bayesian tutorial

  1996influential reference
• Fast Exact Multiplication by the Hessian

  1994cited by this paper
• Signal detection theory and psychophysics

  1966cited by this paper
• The NIST Year 2012 Speaker Recognition Evaluation Plan 1 I

  year unknowninfluential reference
• The NIST Year 2010 Speaker Recognition Evaluation Plan

  year unknowninfluential reference

• A Copula-Based Generative Score-Level Fusion Model for Speaker Verification

  2025cites this paper
• Self-Tuning Spectral Clustering for Speaker Diarization

  2024cites this paper
• A Generative Approach to Condition-Aware Score Calibration for Speaker Verification

  2023cites this paper
• The Distributions of Uncalibrated Speaker Verification Scores: A Generative Model for Domain Mismatch and Trial-Dependent Calibration

  2023cites this paper
• Analysis and Comparison of Classification Metrics

  2022cites this paper
• Impostor Score Statistics as Quality Measures for the Calibration of Speaker Verification Systems

  2022cites this paper
• A Generative Model for Duration-Dependent Score Calibration

  2021cites this paper
• On the Distribution of Speaker Verification Scores: Generative Models for Unsupervised Calibration

  2021influential citation
• Out of a hundred trials, how many errors does your speaker verifier make?

  2021cites this paper
• Deep Learning Models

  2020cites this paper
• Machine Learning for Speaker Recognition

  2020cites this paper
• On Open-Set Speaker Identification with I-Vectors

  2020cites this paper
• Probabilistic embeddings for speaker diarization

  2020cites this paper
• Normal Variance-Mean Mixtures for Unsupervised Score Calibration

  2019influential citation
• Tied Normal Variance–Mean Mixtures for Linear Score Calibration

  2019cites this paper
• The Usefulness of Imperfect Speech Data for ASR Development in Low-Resource Languages

  2019cites this paper
• A probabilistic approach to quantify the strength of evidence of presence of cell types from RNA data using a multi-label method

  2019influential citation
• DNN-Based Score Calibration With Multitask Learning for Noise Robust Speaker Verification

  2018cites this paper
• Leveraging side information for speaker identification with the Enron conversational telephone speech collection

  2017cites this paper
• EVALUATION OF SCIENTIFIC EVIDENCE : A PROPOSAL ON ONTOLOGICAL AND EPISTEMOLOGICAL BASES, AND SOME STATISTICAL APPLICATIONS

  2017cites this paper
• Analysis of the Impact of the Audio Database Characteristics in the Accuracy of a Speaker Clustering System

  2016cites this paper
• Effective speaker spotting for watch-list detection of fraudsters in telephone banking

  2015cites this paper
• Modified-prior PLDA and score calibration for duration mismatch compensation in speaker recognition system

  2015cites this paper
• Test % Data % Out % of , Domain % Labeled % Data % Segmentation

  2014cites this paper
• A comparison of linear and non-linear calibrations for speaker recognition

  2014influential citation
• Speaker diarization with plda i-vector scoring and unsupervised calibration

  2014cites this paper
• Improving speaker recognition performance in the domain adaptation challenge using deep neural networks

  2014cites this paper
• Unsupervised Domain Adaptation for I-Vector Speaker Recognition

  2014cites this paper
• Bayesian calibration for forensic evidence reporting

  2014cites this paper
• Unsupervised Training of PLDA with Variational Bayes

  2014cites this paper
• Robust speaker verification

  2014cites this paper