A One-Pass Real-Time Decoder Using Memory-Efficient State Network
スポンサーリンク
概要
- 論文の詳細を見る
This paper presents our developed decoder which adopts the idea of statically optimizing part of the knowledge sources while handling the others dynamically. The lexicon, phonetic contexts and acoustic model are statically integrated to form a memory-efficient state network, while the language model (LM) is dynamically incorporated on the fly by means of extended tokens. The novelties of our approach for constructing the state network are (1) introducing two layers of dummy nodes to cluster the cross-word (CW) context dependent fan-in and fan-out triphones, (2) introducing a so-called “WI layer” to store the word identities and putting the nodes of this layer in the non-shared mid-part of the network, (3) optimizing the network at state level by a sufficient forward and backward node-merge process. The state network is organized as a multi-layer structure for distinct token propagation at each layer. By exploiting the characteristics of the state network, several techniques including LM look-ahead, LM cache and beam pruning are specially designed for search efficiency. Especially in beam pruning, a layer-dependent pruning method is proposed to further reduce the search space. The layer-dependent pruning takes account of the neck-like characteristics of WI layer and the reduced variety of word endings, which enables tighter beam without introducing much search errors. In addition, other techniques including LM compression, lattice-based bookkeeping and lattice garbage collection are also employed to reduce the memory requirements. Experiments are carried out on a Mandarin spontaneous speech recognition task where the decoder involves a trigram LM and CW triphone models. A comparison with HDecode of HTK toolkits shows that, within 1% performance deviation, our decoder can run 5 times faster with half of the memory footprint.
- （社）電子情報通信学会の論文
- 2008-03-01
著者
-
Yan Yonghong
Thinkit Speech Lab. Institute Of Acoustics Chinese Academy Of Sciences
-
Yan Yonghong
Institute Of Acoustics Chinese Academy Of Science
-
Zhao Qingwei
Thinkit Speech Lab Institute Of Acoustics Chinese Academy Of Sciences
-
Yan Yonghong
Thinkit Speech Lab Institute Of Acoustics Chinese Academy Of Sciences
-
Yan Yonghong
Thinkit Speech Lab.
-
Yan Yonghong
Thinkit Speech Laboratory Institute Of Acoustics Chinese Academy Of Sciences Beijing
-
SHAO Jian
ThinkIT Speech Lab.
-
LI Ta
ThinkIT Speech Lab.
-
ZHANG Qingqing
ThinkIT Speech Lab.
-
Shao Jian
Thinkit Speech Lab Institute Of Acoustics Chinese Academy Of Sciences
-
Zhang Qingqing
Thinkit Speech Laboratory Institute Of Acoustics Chinese Academy Of Sciences Beijing
関連論文
- Effects of single-channel speech enhancement algorithms on Mandarin speech intelligibility (応用音響)
- Approximate Decision Function and Optimization for GMM-UBM Based Speaker Verification
- Using a Kind of Novel Phonotactic Information for SVM Based Speaker Recognition
- Robust Speaker Clustering Using Affinity Propagation
- An LVCSR Based Reading Miscue Detection System Using Knowledge of Reference and Error Patterns
- Effective Acoustic Modeling for Pronunciation Quality Scoring of Strongly Accented Mandarin Speech
- A One-Pass Real-Time Decoder Using Memory-Efficient State Network
- Development of a Mandarin-English Bilingual Speech Recognition System for Real World Music Retrieval
- Automatic Singing Performance Evaluation for Untrained Singers
- Melody Track Selection Using Discriminative Language Model
- Automatic Language Identification with Discriminative Language Characterization Based on SVM
- A two-element-microphone-array-based speech recognition system in vehicle environment(Commemoration of the Japan-China Joint Conference on Acoustics 2007 (JCA2007))
- Speech Enhancement Using Improved Adaptive Null-Forming in Frequency Domain with Postfilter
- Effects of the Temporal Fine Structure in Different Frequency Bands on Mandarin Tone Perception
- Acoustic Feature Optimization Based on F-Ratio for Robust Speech Recognition
- A Hybrid Speech Emotion Recognition System Based on Spectral and Prosodic Features
- Enhancing the Robustness of the Posterior-Based Confidence Measures Using Entropy Information for Speech Recognition
- Two-Microphone Noise Reduction Using Spatial Information-Based Spectral Amplitude Estimation
- A bayesian logistic regression approach to spoken language identification