Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 93 === In this thesis, we deal with the issues of model uncertainty and model discriminability when building Bayesian classification rule for large vocabulary continuous speech recognition. In conventional Bayeisan classification, we optimize the criterion of minimum...

Full description

Bibliographic Details
Main Authors: Yu-Chien Weng, 翁毓謙
Other Authors: Jen-Tzung Chien
Format: Others
Language:zh-TW
Published: 2005
Online Access:http://ndltd.ncl.edu.tw/handle/29504570124506789553
id ndltd-TW-093NCKU5392050
record_format oai_dc
spelling ndltd-TW-093NCKU53920502017-08-27T04:29:40Z http://ndltd.ncl.edu.tw/handle/29504570124506789553 Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition 鑑別性貝氏分類法則應用於大詞彙連續語音辨識 Yu-Chien Weng 翁毓謙 碩士 國立成功大學 資訊工程學系碩博士班 93 In this thesis, we deal with the issues of model uncertainty and model discriminability when building Bayesian classification rule for large vocabulary continuous speech recognition. In conventional Bayeisan classification, we optimize the criterion of minimum Bayes risk (MBR) where the zero-one loss function is considered. The resulting maximum a posteriori (MAP) classification rule has been applied in many speech recognition systems. To improve discriminability of pattern classifier, it is important to design a discriminative loss function where input speech classified to different models should be properly penalized. In this study, we develop a Bayes factor based loss function. This loss/penalty function is established by performing hypothesis test of input speech corresponding to a target model against a competing model. The predictive distributions of target and competing models are computed to determine Bayes factors. In general, the new classification rule is discriminative and robust since the competing model and parameter uncertainty are considered in loss function. We also realize the proposed discriminative Bayesian classification in word graph based search algorithm. From the estimated word candidates and corresponding states, we can calculate loss functions and used them for word graph rescoring of individual word candidates. In the evaluation of broadcast news transcription using MATBN database, we show the superiority of proposed classification compared to MAP classification. Jen-Tzung Chien 簡仁宗 2005 學位論文 ; thesis 93 zh-TW
collection NDLTD
language zh-TW
format Others
sources NDLTD
description 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 93 === In this thesis, we deal with the issues of model uncertainty and model discriminability when building Bayesian classification rule for large vocabulary continuous speech recognition. In conventional Bayeisan classification, we optimize the criterion of minimum Bayes risk (MBR) where the zero-one loss function is considered. The resulting maximum a posteriori (MAP) classification rule has been applied in many speech recognition systems. To improve discriminability of pattern classifier, it is important to design a discriminative loss function where input speech classified to different models should be properly penalized. In this study, we develop a Bayes factor based loss function. This loss/penalty function is established by performing hypothesis test of input speech corresponding to a target model against a competing model. The predictive distributions of target and competing models are computed to determine Bayes factors. In general, the new classification rule is discriminative and robust since the competing model and parameter uncertainty are considered in loss function. We also realize the proposed discriminative Bayesian classification in word graph based search algorithm. From the estimated word candidates and corresponding states, we can calculate loss functions and used them for word graph rescoring of individual word candidates. In the evaluation of broadcast news transcription using MATBN database, we show the superiority of proposed classification compared to MAP classification.
author2 Jen-Tzung Chien
author_facet Jen-Tzung Chien
Yu-Chien Weng
翁毓謙
author Yu-Chien Weng
翁毓謙
spellingShingle Yu-Chien Weng
翁毓謙
Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition
author_sort Yu-Chien Weng
title Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition
title_short Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition
title_full Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition
title_fullStr Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition
title_full_unstemmed Discriminative Bayesian Classification for Large Vocabulary Continuous Speech Recognition
title_sort discriminative bayesian classification for large vocabulary continuous speech recognition
publishDate 2005
url http://ndltd.ncl.edu.tw/handle/29504570124506789553
work_keys_str_mv AT yuchienweng discriminativebayesianclassificationforlargevocabularycontinuousspeechrecognition
AT wēngyùqiān discriminativebayesianclassificationforlargevocabularycontinuousspeechrecognition
AT yuchienweng jiànbiéxìngbèishìfēnlèifǎzéyīngyòngyúdàcíhuìliánxùyǔyīnbiànshí
AT wēngyùqiān jiànbiéxìngbèishìfēnlèifǎzéyīngyòngyúdàcíhuìliánxùyǔyīnbiànshí
_version_ 1718518489403097088