Eigenvoice modeling with sparse training data

Kenny, Patrick; Boulianne, Gilles; Dumouchel, Pierre

doi:10.1109/tsa.2004.840940

Cited by 421 publications

(295 citation statements)

References 17 publications

Supporting

Mentioning

289

Contrasting

Unclassified

Order By: Relevance

“…Compared to traditional i-vector extraction [4], the S (c) filtering mechanism ensure that only those key components representing lexical information of utterance involve in adaptation. We use α = R/C to denote the ratio of the number of key components to the number of total components.…”

Section: Text-dependent I-vector Extractionmentioning

confidence: 99%

See 1 more Smart Citation

Lexicon-Based Local Representation for Text-Dependent Speaker Verification

You

Zhu

2017

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

SUMMARYA text-dependent i-vector extraction scheme and a lexicon-based binary vector (L-vector) representation are proposed to improve the performance of text-dependent speaker verification. I-vector and L-vector are used to represent the utterances for enrollment and test. An improved cosine distance kernel is constructed by combining i-vector and L-vector together and is used to distinguish both speaker identity and lexical (or text) diversity with back-end support vector machine (SVM). Experiments are conducted on RSR 2015 Corpus part 1 and part 2, the results indicate that at most 30% improvement can be obtained compared with traditional i-vector baseline.

show abstract

Section: Text-dependent I-vector Extractionmentioning

confidence: 99%

“…An improved cosine distance kernel is constructed in Sect. 4. Experiments are conducted and several results are presented and analysed in Sect.…”

Section: Introductionmentioning

confidence: 99%

Lexicon-Based Local Representation for Text-Dependent Speaker Verification

You

Zhu

2017

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

show abstract

“…This algorithm is similar to the one used to estimate the identity (betweenclass) subspace in JFA [32], with one major difference: while JFA jointly consider the samples coming from a given subject, TV treats them as if they have been produced by different identities, which is an advantage when large unlabelled training datasets are used. In addition, the extraction of i-vectors requires the estimation of a covariance matrix Σ T to model the residual variability that is not captured by the subspace T .…”

Section: Total Variability Modellingmentioning

confidence: 99%

Bi-modal biometric authentication on mobile phones in challenging conditions

Khoury

Shafey

McCool

et al. 2014

Image and Vision Computing

View full text Add to dashboard Cite

This paper examines the issue of face, speaker and bi-modal authentication in mobile environments when there is significant condition mismatch. We introduce this mismatch by enrolling client models on high quality biometric samples obtained on a laptop computer and authenticating them on lower quality biometric samples acquired with a mobile phone. To perform these experiments we develop three novel authentication protocols for the large publicly available MOBIO database. We evaluate state-of-the-art face, speaker and bi-modal authentication techniques and show that inter-session variability modelling using Gaussian mixture models provides a consistently robust system for face, speaker and bi-modal authentication. It is also shown that multi-algorithm fusion provides a consistent performance improvement for face, speaker and bi-modal authentication. Using this bi-modal multi-algorithm system we derive a state-of-the-art authentication system that obtains a half total error rate of 6.3% and 1.9% for Female and Male trials, respectively.

show abstract

“…There have been reported work on speech [9] and speaker recognition [10] where researchers leverage on existing speech corpora from non-target speakers as the prior knowledge to improve their systems' performance. Following the same idea, eigenvoice-based conversion [11], and tensor representation of speaker space [12] are examples of similar successful attempts in voice conversion.…”

Section: Introductionmentioning

confidence: 99%

Mixture of Factor Analyzers Using Priors From Non-Parallel Speech for Voice Conversion

Kinnunen

Chng

et al. 2012

IEEE Signal Process. Lett.

View full text Add to dashboard Cite

Abstract-A robust voice conversion function relies on a large amount of parallel training data, which is difficult to collect in practice. To tackle the sparse parallel training data problem in voice conversion, this paper describes a mixture of factor analyzers method which integrates prior knowledge from nonparallel speech into the training of conversion function. The experiments on CMU ARCTIC corpus show that the proposed method improves the quality and similarity of converted speech. With both objective and subjective evaluations, we show the proposed method outperforms the baseline GMM method.Index Terms-Voice conversion, prior knowledge, factor analysis, mixture of factor analyzers.

show abstract

Eigenvoice modeling with sparse training data

Cited by 421 publications

References 17 publications

Lexicon-Based Local Representation for Text-Dependent Speaker Verification

Lexicon-Based Local Representation for Text-Dependent Speaker Verification

Bi-modal biometric authentication on mobile phones in challenging conditions

Mixture of Factor Analyzers Using Priors From Non-Parallel Speech for Voice Conversion

Contact Info

Product

Resources

About