Eigenvoice modeling with sparse training data
IEEE Transactions on Speech and Audio ProcessingPublished 19 April 2005
Patrick Kenny, Gilles Boulianne, Pierre Dumouchel
Citations472
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
This work derives an exact solution to the problem of maximum likelihood estimation of the supervector covariance matrix used in extended MAP (or EMAP) speaker adaptation and shows how it can be regarded as a new method of eigenvoice estimation.
Abstract
We derive an exact solution to the problem of maximum likelihood estimation of the supervector covariance matrix used in extended MAP (or EMAP) speaker adaptation and show how it can be regarded as a new method of eigenvoice estimation. Unlike other approaches to the problem of estimating eigenvoices in situations where speaker-dependent training is not feasible, our method enables us to estimate as many eigenvoices from a given training set as there are training speakers. In the limit as the amount of training data for each speaker tends to infinity, it is equivalent to cluster adaptive training.
Keywords
Computer ScienceEngineering
Neural ComputationMixtures of Probabilistic Principal Component Analyzers
1,903 Citations1999Michael E. Tipping, Chris Bishop
PCA is formulated within a maximum likelihood framework, based on a specific form of gaussian latent variable model, which leads to a well-defined mixture model for probabilistic principal component analyzers, whose parameters can be determined using an expectation-maximization algorithm.
PsychometrikaEM Algorithms for ML Factor Analysis
602 Citations1982Donald B. Rubin, Dorothy T. Thayer
Probability, Random Processes, and Estimation Theory for Engineers
593 Citations1986Henry Stark, John W. Woods
IEEE Transactions on Speech and Audio ProcessingSemi-tied covariance matrices for hidden Markov models
538 Citations1999Mark Gales
A new form of covariance matrix which allows a few "full" covariance matrices to be shared over many distributions, whilst each distribution maintains its own "diagonal" covariancy matrix is introduced.
IEEE Transactions on Speech and Audio ProcessingRapid speaker adaptation in eigenvoice space
487 Citations2000Roland Kühn, J.-C. Junqua +2 more
A new model-based speaker adaptation algorithm called the eigenvoice approach, which constrains the adapted model to be a linear combination of a small number of basis vectors obtained offline from a set of reference speakers, and thus greatly reduces the number of free parameters to be estimated from adaptation data.
IEEE Transactions on Speech and Audio ProcessingCluster adaptive training of hidden Markov models
253 Citations2000Mark Gales
This paper examines an adaptation scheme requiring very few parameters, cluster adaptive training (CAT), which may be viewed as a simple extension to speaker clustering, a linear interpolation of all the cluster means is used as the mean of the particular speaker.
Applied and Computational Harmonic AnalysisRandom Cascades on Wavelet Trees and Their Use in Analyzing and Modeling Natural Images
200 Citations2001Martin J. Wainwright, Eero P. Simoncelli +1 more
Proceedings of the IEEEOn adaptive decision rules and decision parameter adaptation for automatic speech recognition
130 Citations2000Chin‐Hui Lee, Qiang Huo
The mathematical framework for Bayesian adaptation of acoustic and language model parameters is first described, and maximum a posteriori point estimation is developed for hidden Markov models and a number of useful parameters densities commonly used in automatic speech recognition and natural language processing.
Batch, incremental and instantaneous adaptation techniques for speech recognition
89 Citations2002G. Zavaliagkos, Richard Schwartz +1 more
It is shown that sizable gains can be achieved by either batch or incremental adaptation for large vocabulary recognition of native speakers, and that good improvements in performance are realized when instantaneous adaptation is used for recognition of non-native speakers.
New MAP estimators for speaker recognition
56 Citations2003Patrick Kenny, M. Mihoubi +1 more
It is shown how eigenvoice MAP can be modified to yield a new model-based channel compensation technique which is called eigenchannel MAP, which was found to reduce speaker identification errors by 50%.
Separation of speech signals using iterative multi-pitch analysis and prediction
48 Citations1999Matti Karjalainen, Tero Tolonen
IEEE Transactions on Speech and Audio ProcessingOn-line adaptive learning of the correlated continuous density hidden Markov models for speech recognition
37 Citations1998Qiang Huo, Chin‐Hui Lee
IEEE Transactions on Speech and Audio ProcessingA Markov random field approach to Bayesian speaker adaptation
28 Citations1997B.M. Shahshahani
Fast speaker adaptation using a priori knowledge
25 Citations1999Roland Kühn, P. Nguyen +6 more
How to find the eigenvoices, give a maximum-likelihood estimator for the new speaker's eigenvoice coefficients, and summarize mean adaptation experiments carried out on the Isolet database are reviewed.
Perspectives in neural computingHidden Markov Independent Component Analysis
24 Citations2000W.D. Penny, Richard Everson +1 more
A generative model is proposed for the analysis of non-stationary multivariate time series that uses a hidden Markov process to switch between independent component models where the components themselves are modelled as generalised autoregressive processes.
Eigenfaces and eigenvoices: dimensionality reduction for specialized pattern recognition
23 Citations2002Roland Kühn, P. Nguyen +2 more
It is argued that speech researchers should use similar techniques to represent variation between speakers, and applications to speaker adaptation, speaker identification and speaker verification are discussed.
Anisotropic MAP defined by eigenvoices for large vocabulary continuous speech recognition
17 Citations2002Henrik Botterweck
A general method is examined, which unifies the eigenvoice approach and MAP adaptation, and has two advantages: long term adaptation leads to the same good results as the MAP method, whereas for ultra-short adaptation in the range of 1-2 seconds an overfitting as for maximum likelihood techniques is avoided.
French large vocabulary recognition with cross-word phonology transducers
15 Citations2002Gilles Boulianne, Julie Brousseau +2 more
This work describes the implementation of French large vocabulary recognition based on transducers, and how it takes advantage of this approach to integrate automatic pronunciation rules and cross-word phenomena such as French "liaison".
IEEE Signal Processing LettersRobust correlation estimation for EMAP-based speaker adaptation
11 Citations2001Eugene Jon, Dong Kook Kim +1 more
Through various experiments on continuous digit recognition, it is shown that the EMAP approach based on the PPCA gives enhanced performance, especially for a small amount of adaptation data.
EMAP-based speaker adaptation with robust correlation estimation
8 Citations2002Eugene Jon, Dong Kook Kim +1 more
Through various experiments on continuous digit recognition, it is shown that the EMAP approach based on the PPCA gives enhanced performance especially for a small amount of adaptation data.