Speaker adaptation of continuous density HMMs using multivariate linear regression

Leggetter, C. J.; Woodland, Philip C.

doi:10.21437/icslp.1994-127

Cited by 63 publications

(10 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The translation of all words is the indefinite article a. But there are also highly confusable words with different meanings, for example Maus , Haus , Klaus 1 , or M u t , W u t , H u t 2 , or erlangen , ergangen 3 .…”

Section: Databasementioning

confidence: 99%

See 1 more Smart Citation

On the influence of hyperarticulated speech on recognition performance

Soltau¹,

Waibel²

1998

5th International Conference on Spoken Language Processing (ICSLP 1998)

View full text Add to dashboard Cite

Since we cannot exclude that speech recognizers fail sometimes, it is important to examine how users react to recognition errors. In correction situations, speaking style becomes more accentuated to disambiguate the original mistake. We examine the effect of speaking style in such situations on speech recognition performance. Our results indicate that hyperarticulated effects occur in correction situations and decrease word accuracy significantly.

show abstract

Section: Databasementioning

confidence: 99%

“…We examined two methods to adapt acoustic models. First, the codebooks were supervised adapted with MLLR [3]. For the adaptation we used only the normally spoken data.…”

Section: Acoustic and Transition Modelsmentioning

confidence: 99%

On the influence of hyperarticulated speech on recognition performance

Soltau¹,

Waibel²

1998

5th International Conference on Spoken Language Processing (ICSLP 1998)

View full text Add to dashboard Cite

show abstract

“…In general, they aect the features of speech in a non-linear fashion which is not easily characterized. Most approaches either adapt the features [1,2,3] or the model parameters [4,5]. Recently, stochastic matching, a framework for adapting both the features and the models using the maximum likelihood approach w as developed [6].…”

Section: Introductionmentioning

confidence: 99%

Maximum-likelihood stochastic matching approach to non-linear equalization for robust speech recognition

Surendran¹,

Lee²,

Rahim³

1996

4th International Conference on Spoken Language Processing (ICSLP 1996)

View full text Add to dashboard Cite

In this paper we present a new technique in the stochastic matching framework to compensate for non-linear distortions in speech recognition. The features of the test data and the means of the trained model are both transformed using neural networks to better t each other. The parameters of the neural network are estimated using a novel combination of the generalized EM (GEM) and the backpropagation algorithms. In the feature transformation case, when the exact Q-functions cannot be calculated, approximations are heuristically derived. The mathematical properties of the new algorithm are analyzed. The performance of the algorithm is also studied under dierent mismatch conditions.

show abstract

“…Second, the amount of calibration data is less than the amount of training data used to estimate the SI parameters, and therefore it is only possible to robustly estimate fewer parameters. In response to this problem, parameter tying and linear regression techniques have been used where the large number of parameters are mapped to a smaller space by a simple transformation [5,9]. This leads to a second problem: there exists no simple transformation from the parameters of the SI system to an efficient model of the speaker variation.…”

Section: Introductionmentioning

confidence: 99%

Speaker adaptation by modeling the speaker variation in a continuous speech recognition system

Ström¹

1996

4th International Conference on Spoken Language Processing (ICSLP 1996)

View full text Add to dashboard Cite

A method for unsupervised instantaneous speaker adaptation is presented and evaluated on a continuous speech recognition task in a man-machine dialogue system. The method is based on modeling of the systematic speaker variation. The variation is modeled by a low-dimensional speaker space and the classification of speech segments is conditioned by the position in the speaker space. Because the effect of the speaker space position on the classification is determined in an off-line training procedure using the speakers in a training database, complex systematic speaker variation can be modeled. Speaker adaptation is achieved only by the constraint that the position in the speaker space is constant over each utterance. Therefore, no separate adaptation session is needed and the adaptation is present from the first utterance. Consequently, for a user there is no noticeable difference between this system and a speaker-independent system. The speaker model and the phonetic classification are implemented in the ANN part of a hybrid ANN/HMM system. In experiments with a pilot system, word accuracy is improved for utterances longer than three words and utterance level results are improved for utterances of all lengths.

show abstract

Speaker adaptation of continuous density HMMs using multivariate linear regression

Cited by 63 publications

References 0 publications

On the influence of hyperarticulated speech on recognition performance

On the influence of hyperarticulated speech on recognition performance

Maximum-likelihood stochastic matching approach to non-linear equalization for robust speech recognition

Speaker adaptation by modeling the speaker variation in a continuous speech recognition system

Contact Info

Product

Resources

About