Study of senone-based deep neural network approaches for spoken language recognition

This paper compares different approaches for using deep neural networks (DNNs) trained to predict senone posteriors for the task of spoken language recognition (SLR). These approaches have recently been found to outperform various baseline systems on different datasets, but they have not yet been co...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Ferrer, L., Lei, Y., McLaren, M., Scheffer, N.
Formato: JOUR
Materias:
Deep neural networks (DNNs)
Senones
Spoken language recognition (SLR)
Forecasting
Gaussian distribution
Speech recognition
Vectors
Bottleneck features
Deep neural networks
Gaussian Mixture Model
Language identification
Language recognition
Score-level fusion
Spoken language recognition
Computational linguistics
Acceso en línea:http://hdl.handle.net/20.500.12110/paper_23299290_v24_n1_p105_Ferrer
Aporte de:
Biblioteca Digital - Facultad de Ciencias Exactas y Naturales (UBA) de Universidad de Buenos Aires
id todo:paper_23299290_v24_n1_p105_Ferrer
record_format dspace
spelling todo:paper_23299290_v24_n1_p105_Ferrer2023-10-03T16:41:01Z Study of senone-based deep neural network approaches for spoken language recognition Ferrer, L. Lei, Y. McLaren, M. Scheffer, N. Deep neural networks (DNNs) Senones Spoken language recognition (SLR) Forecasting Gaussian distribution Speech recognition Vectors Bottleneck features Deep neural networks Gaussian Mixture Model Language identification Language recognition Score-level fusion Senones Spoken language recognition Computational linguistics This paper compares different approaches for using deep neural networks (DNNs) trained to predict senone posteriors for the task of spoken language recognition (SLR). These approaches have recently been found to outperform various baseline systems on different datasets, but they have not yet been compared to each other or to a common baseline. Two of these approaches use the DNNs to generate feature vectors which are then processed in different ways to predict the score of each language given a test sample. The features are extracted either from a bottleneck layer in the DNN or from the output layer. In the third approach, the standard i-vector extraction procedure is modified to use the senones as classes and the DNN to predict the zeroth order statistics. We compare these three approaches and conclude that the approach based on bottleneck features followed by i-vector modeling outperform the other two approaches. We also show that score-level fusion of some of these approaches leads to gains over using a single approach for short-duration test samples. Finally, we demonstrate that fusing systems that use DNNs trained with several languages leads to improvements in performance over the best single system, and we propose an adaptation procedure for DNNs trained with languages with less available data. Overall, we show improvements between 40% and 70% relative to a state-of-the-art Gaussian mixture model (GMM) i-vector system on test durations from 3 seconds to 120 seconds on two significantly different tasks: the NIST 2009 language recognition evaluation task and the DARPA RATS language identification task. © 2015 IEEE. JOUR info:eu-repo/semantics/openAccess http://creativecommons.org/licenses/by/2.5/ar http://hdl.handle.net/20.500.12110/paper_23299290_v24_n1_p105_Ferrer
institution Universidad de Buenos Aires
institution_str I-28
repository_str R-134
collection Biblioteca Digital - Facultad de Ciencias Exactas y Naturales (UBA)
topic Deep neural networks (DNNs)
Senones
Spoken language recognition (SLR)
Forecasting
Gaussian distribution
Speech recognition
Vectors
Bottleneck features
Deep neural networks
Gaussian Mixture Model
Language identification
Language recognition
Score-level fusion
Senones
Spoken language recognition
Computational linguistics
spellingShingle Deep neural networks (DNNs)
Senones
Spoken language recognition (SLR)
Forecasting
Gaussian distribution
Speech recognition
Vectors
Bottleneck features
Deep neural networks
Gaussian Mixture Model
Language identification
Language recognition
Score-level fusion
Senones
Spoken language recognition
Computational linguistics
Ferrer, L.
Lei, Y.
McLaren, M.
Scheffer, N.
Study of senone-based deep neural network approaches for spoken language recognition
topic_facet Deep neural networks (DNNs)
Senones
Spoken language recognition (SLR)
Forecasting
Gaussian distribution
Speech recognition
Vectors
Bottleneck features
Deep neural networks
Gaussian Mixture Model
Language identification
Language recognition
Score-level fusion
Senones
Spoken language recognition
Computational linguistics
description This paper compares different approaches for using deep neural networks (DNNs) trained to predict senone posteriors for the task of spoken language recognition (SLR). These approaches have recently been found to outperform various baseline systems on different datasets, but they have not yet been compared to each other or to a common baseline. Two of these approaches use the DNNs to generate feature vectors which are then processed in different ways to predict the score of each language given a test sample. The features are extracted either from a bottleneck layer in the DNN or from the output layer. In the third approach, the standard i-vector extraction procedure is modified to use the senones as classes and the DNN to predict the zeroth order statistics. We compare these three approaches and conclude that the approach based on bottleneck features followed by i-vector modeling outperform the other two approaches. We also show that score-level fusion of some of these approaches leads to gains over using a single approach for short-duration test samples. Finally, we demonstrate that fusing systems that use DNNs trained with several languages leads to improvements in performance over the best single system, and we propose an adaptation procedure for DNNs trained with languages with less available data. Overall, we show improvements between 40% and 70% relative to a state-of-the-art Gaussian mixture model (GMM) i-vector system on test durations from 3 seconds to 120 seconds on two significantly different tasks: the NIST 2009 language recognition evaluation task and the DARPA RATS language identification task. © 2015 IEEE.
format JOUR
author Ferrer, L.
Lei, Y.
McLaren, M.
Scheffer, N.
author_facet Ferrer, L.
Lei, Y.
McLaren, M.
Scheffer, N.
author_sort Ferrer, L.
title Study of senone-based deep neural network approaches for spoken language recognition
title_short Study of senone-based deep neural network approaches for spoken language recognition
title_full Study of senone-based deep neural network approaches for spoken language recognition
title_fullStr Study of senone-based deep neural network approaches for spoken language recognition
title_full_unstemmed Study of senone-based deep neural network approaches for spoken language recognition
title_sort study of senone-based deep neural network approaches for spoken language recognition
url http://hdl.handle.net/20.500.12110/paper_23299290_v24_n1_p105_Ferrer
work_keys_str_mv AT ferrerl studyofsenonebaseddeepneuralnetworkapproachesforspokenlanguagerecognition
AT leiy studyofsenonebaseddeepneuralnetworkapproachesforspokenlanguagerecognition
AT mclarenm studyofsenonebaseddeepneuralnetworkapproachesforspokenlanguagerecognition
AT scheffern studyofsenonebaseddeepneuralnetworkapproachesforspokenlanguagerecognition
_version_ 1807323551707955200

Ejemplares similares