Speaker recognition with session variability normalization based on MLLR adaptation transforms

DMCA

Speaker recognition with session variability normalization based on MLLR adaptation transforms (1987)

Cached

Download Links

by Andreas Stolcke , Senior Member , Sachin S. Kajarekar , Luciana Ferrer , Elizabeth Shriberg

Venue:	IEEE Trans. Audio, Speech, and Lang. Process
Citations:	29 - 7 self

BibTeX

@ARTICLE{Stolcke87speakerrecognition,
    author = {Andreas Stolcke and Senior Member and Sachin S. Kajarekar and Luciana Ferrer and Elizabeth Shriberg},
    title = {Speaker recognition with session variability normalization based on MLLR adaptation transforms},
    journal = {IEEE Trans. Audio, Speech, and Lang. Process},
    year = {1987},
    volume = {15}
}

OpenURL

Abstract

Abstract—We present a new modeling approach for speaker recognition that uses the maximum-likelihood linear regression (MLLR) adaptation transforms employed by a speech recognition system as features for support vector machine (SVM) speaker models. This approach is attractive because, unlike standard frame-based cepstral speaker recognition models, it normalizes for the choice of spoken words in text-independent speaker verification without data fragmentation. We discuss the basics of the MLLR-SVM approach, and show how it can be enhanced by combining transforms relative to multiple reference models, with excellent results on recent English NIST evaluation sets. We then show how the approach can be applied even if no full word-level recognition system is available, which allows its use on non-English data even without matching speech recognizers. Finally, we examine how two recently proposed algorithms for intersession variability compensation perform in conjunction with MLLR-SVM. Index Terms—Intersession variability compensation, maximum-likelihood linear regression-support vector machine (MLLR-SVM), speaker recognition. I.

Keyphrases

speaker recognition session variability normalization mllr adaptation transforms index term intersession variability compensation maximum-likelihood linear regression-support vector machine adaptation transforms support vector machine non-english data excellent result speech recognition system reference model text-independent speaker verification speech recognizers maximum-likelihood linear regression spoken word full word-level recognition system mllr-svm approach recent english nist evaluation set new modeling approach data fragmentation intersession variability compensation perform speaker model

DMCA