MLP Internal Representation as Discriminative Features for Improved Speaker Recognition

Wu, Dalei; Morris, Andrew; Koreman, Jacques

doi:10.1007/11613107_5

Cited by 17 publications

(16 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The most representative methods include some normalisation methods at the feature level, such as cepstral mean subtraction (CMS), H-norm, etc. (Wu et al, 2008) and transformation techniques at the feature level, such as LDA-based and NLDA-based feature transformation (Wu et al, 2005a(Wu et al, ,b,c, 2008. Processing at the model level then uses a variety of normalisation and transformation techniques at the model level that is a middle stage of a standard recognition system.…”

Section: Introductionmentioning

confidence: 99%

-Gaussian mixture modelling for speaker recognition

2009

Pattern Recognition Letters

Self Cite

View full text Add to dashboard Cite

Section: Introductionmentioning

confidence: 99%

-Gaussian mixture modelling for speaker recognition

2009

Pattern Recognition Letters

Self Cite

View full text Add to dashboard Cite

“…In initial tests with TIMIT in [10] and noisy TINIT in [11 ] we showed that the performance of the features provided by this MLP architecture increases with the number of speakers which the MLP is trained to separate. Due to various factors, test results are subject to significant random fluctuation.…”

Section: Related Researchmentioning

confidence: 99%

MLP trained to separate problem speakers provides improved features for speaker identification

Morris

Koreman

2005

Proceedings 39th Annual 2005 International Carnahan Conference on Security Technology

View full text Add to dashboard Cite

In automatic speech recognition (ASR) the non-linear data prolection provided by a one hidden layer multilayer perceptron (MLP), trained to recognise phonemes, has previously been shown to provide feature enhancement which can substantially increase ASR performance, especially in noise. Previous attempts to apply an analogous approach to speaker identification have not succeeded in improving performance, except by combining MLP processed features with other features. We present test resultsfor the TIMIT database which show that the advantage of MJLP preprocessing for open set speaker identification increases with the number of speakers used to train the MiEP and that improved identification is obtained as this number increases beyond sixty. We also present a methodfor selecting the speakers used for MLEP training which further improves identification performance.

show abstract

“…Instead, we use a subset of speakers, which are most representative of the overall population, and therefore referred to as speaker basis, for MLP training. It is said to be "speaker basis selection" concerning how to select the optimal basis speakers (Wu et al, 2005a(Wu et al, , 2005bMorris et al, 2005). The basic idea for speaker basis selection is based on an approximated Kullback-Leibler (KL) distance between any two speakers, say, S i and S k .…”

Section: Nonlinear Discriminant Analysismentioning

confidence: 99%