A hybrid GMM/SVM approach to speaker identification

Fine, Shai; Navrátil, Jǐŕı; Gopinath, R.

doi:10.1109/icassp.2001.940856

Cited by 68 publications

(35 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this case, we may introduce "slack" variables which represent the amount by which each point is misclassified. In this case, the objective function is reformulated as subject to for all (8) The second term on the right-hand side of (8) is the empirical risk associated with those points that are misclassified or lie within the margin. is a cost function and is a hyper-parameter that trades off the effects of minimizing the empirical risk against maximizing the margin.…”

Section: Support Vector Classificationmentioning

confidence: 99%

“…Score-space kernels include the Fisher kernel [7] and map a complete sequence onto a single point in a high-dimensional space by exploiting generative models. The Fisher kernel has been applied to speaker recognition with limited success [8], [9]. We have applied the score-space kernel SVM approach to text-independent speaker verification, extending some previous work that employed frame-discriminant SVMs [10], [11].…”

mentioning

confidence: 99%

See 1 more Smart Citation

Speaker verification using sequence discriminant support vector machines

Wan

Renals

2005

IEEE Trans. Speech Audio Process.

167

View full text Add to dashboard Cite

Abstract-This paper presents a text-independent speaker verification system using support vector machines (SVMs) with scorespace kernels. Score-space kernels generalize Fisher kernels and are based on underlying generative models such as Gaussian mixture models (GMMs). This approach provides direct discrimination between whole sequences, in contrast with the frame-level approaches at the heart of most current systems. The resultant SVMs have a very high dimensionality since it is related to the number of parameters in the underlying generative model. To address problems that arise in the resultant optimization we introduce a technique called spherical normalization that preconditions the Hessian matrix. We have performed speaker verification experiments using the PolyVar database. The SVM system presented here reduces the relative error rates by 34% compared to a GMM likelihood ratio system. Index Terms-Fisher kernel, score-space kernel, speaker verification, support vector machine.

show abstract

Section: Support Vector Classificationmentioning

confidence: 99%

mentioning

confidence: 99%

Speaker verification using sequence discriminant support vector machines

Wan

Renals

2005

IEEE Trans. Speech Audio Process.

167

View full text Add to dashboard Cite

show abstract

“…Though this paper focuses on speech recognition, it is worth noting that SVMs have already been employed in speaker identification (for example, [9]) and verification (for example, [10]) or to improve confidence measurements that can help in dialog systems [11], among others.…”

Section: Introductionmentioning

confidence: 99%

Robust ASR using Support Vector Machines

Solera-Ureña

Martín-Iglesias

Gallardo-Antolín

et al. 2007

Speech Communication

View full text Add to dashboard Cite

The improved theoretical properties of Support Vector Machines with respect to other machine learning alternatives due to their max-margin training paradigm have led us to suggest them as a good technique for robust speech recognition. However, important shortcomings have had to be circumvented, the most important being the normalisation of the time duration of different realisations of the acoustic speech units.In this paper, we have compared two approaches in noisy environments: first, a hybrid HMM-SVM solution where a fixed number of frames is selected by means of an HMM segmentation and second, a normalisation kernel called Dynamic Time Alignment Kernel (DTAK) first introduced in [1] and based on DTW (Dynamic Time Warping). Special attention has been paid to the adaptation of both alternatives to noisy environments, comparing two types of parameterisations and performing suitable feature normalisation operations. The results show that the DTA Kernel provides important advantages over the baseline HMM system in medium to bad noise conditions, also outperforming the results of the hybrid system.

show abstract

“…The most common approach for current identification systems is based on Gaussian Mixture Models (GMM) [3] or GMMs coupled with Support Vector Machines (GMM-SVM) [4,5]. Verification is accomplished through likelihood comparison with appropriate cohort models, as with a Universal Background Model (UBM) system [6].…”

Section: Introductionmentioning

confidence: 99%