Auditory-based acoustic distinctive features and spectral cues for robust automatic speech recognition in Low-SNR car environments

Selouani, Sid‐Ahmed; Tolba, Hesham; O’Shaughnessy, D.

doi:10.3115/1073483.1073514

Cited by 11 publications

(7 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The obtained results showed that the use of the formant frequencies for ASR in a multi-stream paradigm improves the ASR performance. Then in Selouani et al (2003), we extended our work to evaluate the robustness of the above mentioned proposed features using a multi-stream paradigm for ASR in noisy car environments. The obtained results showed that the use of such features renders the recognition process more robust in noisy car environments.…”

Section: Robust Front-end Processingmentioning

confidence: 99%

“…Another important transformation of the predictor coefficients is the set of partial correlation coefficients or reflection coefficients. In previous papers (Tolba et al 2002;Selouani et al 2003), we introduced a multi-stream paradigm for ASR in which we merge different sources of information about the speech signal that could be lost when using only the MFCCs to recognize uttered speech. Our experiments in Tolba et al (2002) showed that the use of some auditory-based features and formant cues via a multi-stream paradigm approach leads to an improvement of the recognition performance.…”

Section: Robust Front-end Processingmentioning

confidence: 99%

See 1 more Smart Citation

A noise-robust front-end for distributed speech recognition in mobile communications

Addou

Selouani

Kifaya

et al. 2007

Int J Speech Technol

Self Cite

View full text Add to dashboard Cite

This paper investigates a new front-end processing that aims at improving the performance of speech recognition in noisy mobile environments. This approach combines features based on conventional Mel-cepstral Coefficients (MFCCs), Line Spectral Frequencies (LSFs) and formant-like (FL) features to constitute robust multivariate feature vectors. The resulting front-end constitutes an alternative to the DSR-XAFE (XAFE: eXtended Audio FrontEnd) available in GSM mobile communications. Our results showed that for highly noisy speech, using the paradigm that combines these spectral cues leads to a significant improvement in recognition accuracy on the Aurora 2 task.

show abstract

Section: Robust Front-end Processingmentioning

confidence: 99%

Section: Robust Front-end Processingmentioning

confidence: 99%

A noise-robust front-end for distributed speech recognition in mobile communications

Addou

Selouani

Kifaya

et al. 2007

Int J Speech Technol

Self Cite

View full text Add to dashboard Cite

show abstract

“…Another important transformation of the predictor coefficients is the set of partial correlation coefficients or reflection coefficients. In previous papers [2][3][4], we introduced a multi-stream paradigm for ASR in which, we merge different sources of information about the speech signal that could be lost when using only the MFCCs to recognize uttered speech. Our experiments in [2] showed that the use of some auditory-based features and formant cues via a multi-stream paradigm approach leads to an improvement of the recognition performance.…”

Section: Robust Front-end Processingmentioning

confidence: 99%

Feature Combination Using Multiple Spectral Cues for Robust Speech Recognition in Mobile Communications

Addou

Selouani²,

Boudraa

et al. 2009

2009 Sixth International Conference on Information Technology: New Generations

Self Cite

View full text Add to dashboard Cite

This paper investigates a new front-end processing that aims at improving the performance of speech recognition in noisy mobile environments. This approach combines features based on conventional Mel-cepstral Coefficients (MFCCs) and Line Spectral Frequencies (LSFs) to constitute robust multivariate feature vectors. The proposed front-end constitutes an alternative to the DSR-XAFE (XAFE: eXtended Audio Front-End) available in GSM mobile communications. Our results showed that for highly noisy speech, using the paradigm that combines LSF with MFCCs, leads to a significant improvement in recognition accuracy on the Aurora 2 task.

show abstract

“…Cependant, ils précisent qu'il n'est pas nécessaire, pour la majorité des langues, d'utiliser les douze indices. Le modèle auditif calculatoire de Caelen a permis de quantifier ces indices acoustiques, ce qui a ouvert la voie à leur incorporation dans de nombreuses configurations pratiques (Caelen, 1985;Selouani, Tolba et O'Shaughnessy, 2003). • Indice Aigu/Grave (AG) : du point de vue articulatoire, la gravité d'un phonème est générée par un volume plus important de la cavité buccale.…”

Section: Les Indices Acoustiques Auditifs Statiquesunclassified

Un système à base de connaissances pour une communication parlée personne-système multilingue

Selouani

2007

rum

Self Cite

View full text Add to dashboard Cite

La tâche de reconnaissance automatique de la parole (RAP), qui est au coeur de la communication parlée Personne-Système, peut être vue comme une gestion de l’information issue de la microstructure acoustique du signal vocal pour la transformer en une information représentée par la macrostructure phonétique implicite. La correspondance avec le moins d’erreurs possible de ces deux structures nécessite une intégration de connaissances a priori sur la macrostructure phonétique dans des systèmes dédiés à la gestion de l’information acoustico-phonétique. Dans cet article, nous abordons des aspects liés tant à la gestion de l’information phonétique véhiculée par le signal vocal qu’à la topologie de systèmes experts capables de conduire des processus de reconnaissance phonémique multilingue. La démarche que nous proposons consiste à enrichir la base de connaissances de ces experts par des indices représentatifs de la majorité des langues humaines afin de rehausser les performances d’identification des macro-classes et des traits phonétiques divers. Les résultats obtenus sur des corpus de logatomes et de phrases en langues française et arabe montrent qu’il est possible d’orienter la conception des systèmes vers une unification du processus de reconnaissance pour l’adapter à une identification phonémique multilingue.Automatic Speech Recognition (ASR) is at the heart of Man-Machine speech communication. It can be seen as a management of the information emanating from the speech acoustical microstructure. This process aims to transform this information in such a way that it can be represented by the phonetic implicit macrostructure. The effective matching between the two structures requires the integration into expert systems, of an a priori knowledge about the phonetic macrostructures. These expert systems are dedicated to the management of acoustic-phonetic information. This paper investigates aspects linked either to the management of phonetic information contained in the speech signal, or to the topology of expert systems that are capable of conducting a multilingual phonemic recognition process. The proposed method consists of feeding the knowledge base of these expert systems with indicative parameters representing the major human languages in order to enhance the identification performance of phonetic macro-classes and features. The results of experiments carried out on corpora composed of both French and Arabic utterances show that it is possible to conceive systems based on the concept of unified recognition processes dedicated to multilingual phonetic identification

show abstract

Auditory-based acoustic distinctive features and spectral cues for robust automatic speech recognition in Low-SNR car environments

Cited by 11 publications

References 4 publications

A noise-robust front-end for distributed speech recognition in mobile communications

A noise-robust front-end for distributed speech recognition in mobile communications

Feature Combination Using Multiple Spectral Cues for Robust Speech Recognition in Mobile Communications

Un système à base de connaissances pour une communication parlée personne-système multilingue

Contact Info

Product

Resources

About