Fast speaker adaptation of large vocabulary continuous density HMM speech recognizer using a basis transform approach

Boulis, Constantinos; Digalakis, Vassilios

doi:10.1109/icassp.2000.859128

Cited by 6 publications

(4 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An extension of the MLLR technique, where multiple independent transformations are combined, was proposed by Digalakis and colleagues [21,22,46,47]. For the purpose of speaker adaptation on the training material first adaptation transforms are estimated for typical speaker groups.…”

Section: ])mentioning

confidence: 99%

Markov Models for Pattern Recognition

Fink¹

2008

View full text Add to dashboard Cite

Section: ])mentioning

confidence: 99%

Markov Models for Pattern Recognition

Fink¹

2008

View full text Add to dashboard Cite

“…A first version of BT is presented in Boulis and Digalakis (2000). The algorithm was evaluated on the Swedish ATIS corpus where the speaker-independent system is trained on non-dialect speakers.…”

Section: The Basis Transformation Approachmentioning

confidence: 99%

Maximum likelihood stochastic transformation adaptation for medium and small data sets

Boulis

Diakoloukas

Digalakis

2001

Computer Speech & Language

View full text Add to dashboard Cite

Speaker adaptation is recognized as an essential part of today's large-vocabulary automatic speech recognition systems. A family of techniques that has been extensively applied for limited adaptation data is transformation-based adaptation. In transformation-based adaptation we partition our parameter space in a set of classes, estimate a transform (usually linear) for each class and apply the same transform to all the components of the class. It is known, however, that additional gains can be made if we do not constrain the components of each class to use the same transform. In this paper two speaker adaptation algorithms are described. First, instead of estimating one linear transform for each class (as maximum likelihood linear regression (MLLR) does, for example) we estimate multiple linear transforms per class of models and a transform weights vector which is specific to each component (Gaussians in our case). This in effect means that each component receives its own transform without having to estimate each one of them independently. This scheme, termed maximum likelihood stochastic transformation (MLST) achieves a good trade-off between robustness and acoustic resolution. MLST is evaluated on the Wall Street Journal (WSJ) corpus for non-native speakers and it is shown that in the case of 40 adaptation sentences the algorithm outperforms MLLR by more than 13%. In the second half of this paper, we introduce a variant of the MLST designed to operate under sparsity of data. Since the majority of the adaptation parameters are the transformations, we estimate them on the training speakers and adapt to a new speaker by estimating the transform weights only. First we cluster the speakers in a number of sets and estimate the transformations on each cluster. The new speaker will use transformations from all clusters to perform adaptation. This method, termed basis transformation, can be seen as a speaker similarity scheme. Experimental results on the WSJ show that when basis transformation is cascaded with MLLR marginal gains can be obtained from MLLR only, for adaptation of native speakers.

show abstract

“…A closely related problem to nonnative speaker adaptation is regional dialect speaker adaptation. Digalakis et al investigated adapting acoustic models to fit speakers with dialect accents [12], [13]. In [12], Maximum Likelihood Stochastic Transformation (MLST) was proposed to estimate multiple linear transforms for each model cluster in model adaptation.…”

Section: Introductionmentioning

confidence: 99%

“…Although a significant performance improvement was achieved, much more data than that of MLLR were needed, where only one linear transform was estimated for each model cluster in MLLR. In [13], in order to achieve a good performance when adaptation data were sparse, speech data of prototype speakers from target dialect regions were used to generate a set of basis linear transformations and a small amount of new speaker's speech was used to estimate the transform combination weights. In their experiments of Swedish dialect speaker adaptation, the adaptation performance exceeded that of MLLR greatly when the amount of adaptation data was very small.…”

Section: Introductionmentioning

confidence: 99%

Fast model selection based speaker adaptation for nonnative speech

Zhao

2003

IEEE Trans. Speech Audio Process.

View full text Add to dashboard Cite

In this paper, the problem of adapting acoustic models of native English speech to nonnative speakers is addressed from a perspective of adaptive model complexity selection. The goal is to dynamically select model complexity for each nonnative talker so as to optimize the balance between model robustness to pronunciation variations and model detailedness for discrimination of speech sounds. A maximum expected likelihood (MEL) based technique is proposed to enable reliable complexity selection when adaptation data are sparse, where expectation of log-likelihood (EL) of adaptation data is computed based on distributions of mismatch biases between model and data, and model complexity is selected to maximize EL. The MEL based complexity selection is further combined with MLLR to enable adaptation of both complexity and parameters of acoustic models. Experiments were performed on WSJ1 data of speakers with a wide range of foreign accents. Results show that the MEL based complexity selection was feasible when using as little as one adaptation utterance, and it was able to dynamically select proper model complexity as the adaptation data increased. Compared with the standard MLLR, the MEL + MLLR method led to consistent and significant improvement to recognition accuracy on nonnative speakers, without performance degradation on native speakers.

show abstract

Fast speaker adaptation of large vocabulary continuous density HMM speech recognizer using a basis transform approach

Cited by 6 publications

References 6 publications

Markov Models for Pattern Recognition

Markov Models for Pattern Recognition

Maximum likelihood stochastic transformation adaptation for medium and small data sets

Fast model selection based speaker adaptation for nonnative speech

Contact Info

Product

Resources

About