Robust Speech Recognition by Combining Short-Term and Long-Term Spectrum Based Position-Dependent CMN with Conventional CMN

スポンサーリンク

概要

• 論文の詳細を見る

• In a distant-talking environment, the length of channel impulse response is longer than the short-term spectral analysis window. Conventional short-term spectrum based Cepstral Mean Normalization (CMN) is therefore, not effective under these conditions. In this paper, we propose a robust speech recognition method by combining a short-term spectrum based CMN with a long-term one. We assume that a static speech segment (such as a vowel, for example) affected by reverberation, can be modeled by a long-term cepstral analysis. Thus, the effect of long reverberation on a static speech segment may be compensated by the long-term spectrum based CMN. The cepstral distance of neighboring frames is used to discriminate the static speech segment (long-term spectrum) and the non-static speech segment (short-term spectrum). The cepstra of the static and non-static speech segments are normalized by the corresponding cepstral means. In a previous study, we proposed an environmentally robust speech recognition method based on Position-Dependent CMN (PDCMN) to compensate for channel distortion depending on speaker position, and which is more efficient than conventional CMN. In this paper, the concept of combining short-term and long-term spectrum based CMN is extended to PDCMN. We call this Variable Term spectrum based PDCMN (VTPDCMN). Since PDCMN/VT-PDCMN cannot normalize speaker variations because a position-dependent cepstral mean contains the average speaker characteristics over all speakers, we also combine PDCMN/VT-PDCMN with conventional CMN in this study. We conducted the experiments based on our proposed method using limited vocabulary (100 words) distant-talking isolated word recognition in a real environment. The proposed method achieved a relative error reduction rate of 60.9% over the conventional short-term spectrum based CMN and 30.6% over the shortterm spectrum based PDCMN.

• （社）電子情報通信学会の論文
• 2008-03-01

著者

• KITAOKA Norihide

  Nagoya University

• Nakagawa Seiichi

  Toyohashi Univ. Technol. Toyohashi‐shi Jpn

• Nakagawa Seiichi

  Department Of Information And Computer Sciences Toyohashi University

• Nakagawa Seiichi

  Toyohashi Univ. Of Technol. Toyohashi‐shi Jpn

• Kitaoka Norihide

  Toyohashi University Of Technology

• WANG Longbiao

  Toyohashi University of Technology

• Kitaoka Norihide

  Nagoya Univ.

• WANG Longbiao

  Nagoya University

関連論文

• Acoustic Feature Transformation Combining Average and Maximum Classification Error Minimization Criteria
• Acoustic Feature Transformation Based on Discriminant Analysis Preserving Local Structure for Speech Recognition
• Topic dependent language model based on on-line voting (言語理解とコミュニケーション)
• A transitive translation for Indonesian-Japanese CLQA (自然言語処理)
• A Machine Learning Approach for an Indonesian-English Cross Language Question Answering System(Natural Language Processing)
• Indonesian-Japanese Transitive Translation using English for CLIR
• CENSREC-1-C : An evaluation framework for voice activity detection under noisy environments
• Auditory perception versus automatic estimation of location and orientation of an acoustic source in a real environment
• Topic dependent language model based on on-line voting (音声)
• Topic dependent language model based on clustering of noun word history
• Word and class dependency of N-gram language model (音声言語情報処理)
• Word and class dependency of N-gram language model (言語理解とコミュニケーション・第9回音声言語シンポジウム)
• Word and class dependency of N-gram language model (音声・第9回音声言語シンポジウム)
• TEXT-INDEPENDENT SPEAKER IDENTIFICATION ON TIMIT DATABASE
• Text-Independent/Text-Prompted Speaker Recognition by Combining Speaker-Specific GMM with Speaker Adapted Syllable-Based HMM(Speaker Recognition, Statistical Modeling for Speech Processing)
• AURORA-2J: An Evaluation Framework for Japanese Noisy Speech Recognition(Speech Corpora and Related Topics, Corpus-Based Speech Technologies)
• Evaluation of Combinational Use of Discriminant Analysis-Based Acoustic Feature Transformation and Discriminative Training
• LVCSR based on context-dependent syllable acoustic models (Speech) -- (国際ワークショップ"Asian workshop on speech science and technology")
• Robust distant speech recognition by combining variable-term spectrum based position-dependent CMN with conventional CMN (Speech) -- (国際ワークショップ"Asian workshop on speech science and technology")
• Linear Discriminant Analysis Using a Generalized Mean of Class Covariances and Its Application to Speech Recognition
• Robust Speech Recognition by Combining Short-Term and Long-Term Spectrum Based Position-Dependent CMN with Conventional CMN
• LVCSR based on context-dependent syllable acoustic models
• Robust distant speech recognition by combining variable-term spectrum based position-dependent CMN with conventional CMN
• Improving Keyword Recognition of Spoken Queries by Combining Multiple Speech Recognizer's Outputs for Speech-driven WEB Retrieval Task(Spoken Language Systems, Corpus-Based Speech Technologies)
• An Unsupervised Speaker Adaptation Method for Lecture-Style Spontaneous Speech Recognition Using Multiple Recognition Systems(Spoken Language Systems, Corpus-Based Speech Technologies)
• Acoustic Feature Transformation Based on Discriminant Analysis Preserving Local Structure for Speech Recognition
• Speaker Change Detection and Speaker Clustering Using VQ Distortion Measure
• Noisy Speech Recognition Based on Integration/Selection of Multiple Noise Suppression Methods Using Noise GMMs
• Succeeding Word Prediction for Speech Recognition Based on Stochastic Language Model
• Acoustic Feature Transformation Combining Average and Maximum Classification Error Minimization Criteria
• Driver's irritation detection using speech recognition results (音声・第10回音声言語シンポジウム)
• Driver's irritation detection using speech recognition results (音声言語情報処理)
• Driver's irritation detection using speech recognition results (言語理解とコミュニケーション・第10回音声言語シンポジウム)
• A Survey on Automatic Speech Recognition(Special Issue on the 2000 IEICE Excellent Paper Award)
• Relationship among Recognition Rate, Rejection Rate and False Alarm Rate in a Spoken Word Recognition System
• Speaker Recognition by Combining MFCC and Phase Information in Noisy Conditions
• Distant Speech Recognition Using a Microphone Array Network
• Auditory perception versus automatic estimation of location and orientation of an acoustic source in a real environment
• Continuous Speech Recognition Using an On-Line Speaker Adaptation Method Based on Automatic Speaker Clustering (Special Issue on Speech Information Processing)
• Distant-Talking Speech Recognition Based on Spectral Subtraction by Multi-Channel LMS Algorithm
• Response Timing Detection Using Prosodic and Linguistic Information for Human-friendly Spoken Dialog Systems
• A Spoken Dialog System for Spontaneous Conversations Considering Response Timing and Response Type
• INVESTIGATIONS ON TEXT-INDEPENDENT SPEAKER IDENTIFICATION
• Acoustic Model Training Using Pseudo-Speaker Features Generated by MLLR Transformations for Robust Speaker-Independent Speech Recognition
• Selective Gammatone Envelope Feature for Robust Sound Event Recognition
• CENSREC-4: An evaluation framework for distant-talking speech recognition in reverberant environments
• Selective Gammatone Envelope Feature for Robust Sound Event Recognition
• Risk-Based Semi-Supervised Discriminative Language Modeling for Broadcast Transcription
• A Graph-Based Spoken Dialog Strategy Utilizing Multiple Understanding Hypotheses
• Acoustic Model Training Using Pseudo-Speaker Features Generated by MLLR Transformations for Robust Speaker-Independent Speech Recognition
• Risk-Based Semi-Supervised Discriminative Language Modeling for Broadcast Transcription

もっと見る 閉じる

スポンサーリンク