VTLN Using Analytically Determined Linear-Transformation on Conventional MFCC

Sanand, D. Rama; Umesh, S.

doi:10.1109/tasl.2012.2186289

Cited by 8 publications

(8 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is because once VTLN-warping is applied, we have access to L α , but not the power spectrum P. To estimate L α,K from L α , we use the idea of filter-bank interpolation [2].…”

Section: Methodsmentioning

confidence: 99%

“…The authors of [2] proposed filter-bank interpolation for the purpose of estimating L α from L (the unwarped log filter-bank output). Log filter-bank outputs for a sample voiced speech frame from the WSJ database.…”

Section: Methodsmentioning

confidence: 99%

“…Details regarding the derivation of filter-bank interpolation matrices can be found in [2]. To illustrate the combined effect of VTLN and SGR warping, Fig.…”

Section: Methodsmentioning

confidence: 99%

“…The conventional form of vocal-tract length normalization (VTLN) uses a piece-wise linear warping function [1,2], and is quite effective as an unsupervised, utterance-level speaker normalization scheme for automatic speech recognition (ASR) using hidden Markov models (HMMs). However, it is well known that VTLN is not optimal and is sensitive to noise, especially when the training data are clean.…”

Section: Introductionmentioning

confidence: 99%

“…Here, we investigate the combined effect of VTLN and SGR-based warping, and show, using a medium-vocabulary ASR task, that the proposed joint-warping approach provides a significant improvement over VTLN. Our approach is also novel in the way we use filter-bank interpolation (proposed in [2]) to implement VTLN+SGR warping. The Wall Street Journal (WSJ) database and the MFCC front end are used for this part of our study.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Frequency warping using subglottal resonances: Complementarity with VTLN and robustness to additive noise

Arsikere

Alwan

2014

2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

View full text Add to dashboard Cite

Based on our recently-proposed frequency-warping scheme using subglottal resonances (SGRs), this paper addresses two wellknown limitations of conventional vocal-tract length normalization (VTLN): (1) its sub-optimal nature owing to the lack of frequencydependent scaling, and (2) sensitivity to noise. Based on the idea of filter-bank interpolation, a novel approach is proposed to realize the combined effect of VTLN and SGR-based warping (which provides frequency-dependent scaling). Using the Wall Street Journal database and the conventional MFCC front end, SGR warping is shown to be complementary to VTLN in performance. Since SGR warping depends more on the given signal and less on models trained a priori, we argue that SGR warping is less sensitive to noise than VTLN. Through experiments on the AURORA-4 database with power-normalized cepstral coefficients as noise-robust front-end features, we show that SGR warping is better than VTLN, in clean as well as multi-conditional training.

show abstract

“…This is because once VTLN-warping is applied, we have access to L α , but not the power spectrum P. To estimate L α,K from L α , we use the idea of filter-bank interpolation [2].…”

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

“…Details regarding the derivation of filter-bank interpolation matrices can be found in [2]. To illustrate the combined effect of VTLN and SGR warping, Fig.…”

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations