Discriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling

This paper implements the continuous Hindi Automatic Speech Recognition (ASR) system using the proposed integrated features vector with Recurrent Neural Network (RNN) based Language Modeling (LM). The proposed system also implements the speaker adaptation using Maximum-Likelihood Linear Regression (...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Kumar A., Aggarwal R.K.
Formato: article
Lenguaje:EN
Publicado: De Gruyter 2020
Materias:
plp
mmi
mpe
Q
Acceso en línea:https://doaj.org/article/d4a0a539143344e18b0d33ff422f1ec4
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
id oai:doaj.org-article:d4a0a539143344e18b0d33ff422f1ec4
record_format dspace
spelling oai:doaj.org-article:d4a0a539143344e18b0d33ff422f1ec42021-12-05T14:10:51ZDiscriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling2191-026X10.1515/jisys-2018-0417https://doaj.org/article/d4a0a539143344e18b0d33ff422f1ec42020-07-01T00:00:00Zhttps://doi.org/10.1515/jisys-2018-0417https://doaj.org/toc/2191-026XThis paper implements the continuous Hindi Automatic Speech Recognition (ASR) system using the proposed integrated features vector with Recurrent Neural Network (RNN) based Language Modeling (LM). The proposed system also implements the speaker adaptation using Maximum-Likelihood Linear Regression (MLLR) and Constrained Maximum likelihood Linear Regression (C-MLLR). This system is discriminatively trained by Maximum Mutual Information (MMI) and Minimum Phone Error (MPE) techniques with 256 Gaussian mixture per Hidden Markov Model(HMM) state. The training of the baseline system has been done using a phonetically rich Hindi dataset. The results show that discriminative training enhances the baseline system performance by up to 3%. Further improvement of ~7% has been recorded by applying RNN LM. The proposed Hindi ASR system shows significant performance improvement over other current state-of-the-art techniques.Kumar A.Aggarwal R.K.De Gruyterarticleautomatic speech recognitionmfccgfccwerbcplpdiscriminative trainingmmimpernn lmScienceQElectronic computers. Computer scienceQA75.5-76.95ENJournal of Intelligent Systems, Vol 30, Iss 1, Pp 165-179 (2020)
institution DOAJ
collection DOAJ
language EN
topic automatic speech recognition
mfcc
gfcc
werbc
plp
discriminative training
mmi
mpe
rnn lm
Science
Q
Electronic computers. Computer science
QA75.5-76.95
spellingShingle automatic speech recognition
mfcc
gfcc
werbc
plp
discriminative training
mmi
mpe
rnn lm
Science
Q
Electronic computers. Computer science
QA75.5-76.95
Kumar A.
Aggarwal R.K.
Discriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling
description This paper implements the continuous Hindi Automatic Speech Recognition (ASR) system using the proposed integrated features vector with Recurrent Neural Network (RNN) based Language Modeling (LM). The proposed system also implements the speaker adaptation using Maximum-Likelihood Linear Regression (MLLR) and Constrained Maximum likelihood Linear Regression (C-MLLR). This system is discriminatively trained by Maximum Mutual Information (MMI) and Minimum Phone Error (MPE) techniques with 256 Gaussian mixture per Hidden Markov Model(HMM) state. The training of the baseline system has been done using a phonetically rich Hindi dataset. The results show that discriminative training enhances the baseline system performance by up to 3%. Further improvement of ~7% has been recorded by applying RNN LM. The proposed Hindi ASR system shows significant performance improvement over other current state-of-the-art techniques.
format article
author Kumar A.
Aggarwal R.K.
author_facet Kumar A.
Aggarwal R.K.
author_sort Kumar A.
title Discriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling
title_short Discriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling
title_full Discriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling
title_fullStr Discriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling
title_full_unstemmed Discriminatively trained continuous Hindi speech recognition using integrated acoustic features and recurrent neural network language modeling
title_sort discriminatively trained continuous hindi speech recognition using integrated acoustic features and recurrent neural network language modeling
publisher De Gruyter
publishDate 2020
url https://doaj.org/article/d4a0a539143344e18b0d33ff422f1ec4
work_keys_str_mv AT kumara discriminativelytrainedcontinuoushindispeechrecognitionusingintegratedacousticfeaturesandrecurrentneuralnetworklanguagemodeling
AT aggarwalrk discriminativelytrainedcontinuoushindispeechrecognitionusingintegratedacousticfeaturesandrecurrentneuralnetworklanguagemodeling
_version_ 1718371691466326016