Evaluation and Analysis of Minimum Phone Error Training and Its Modified Versions for Large Vocabulary Mandarin Speech Recognition
碩士 === 國立臺灣大學 === 電信工程學研究所 === 96 === The traditional acoustic model training is based on Maximum Likelihood (ML). This training method maximizes the posterior probability of transcription in the training corpus, but cannot guarantee that the incorrect observations do not obtain a larger posterior p...
Main Authors: | Yung-Jen Cheng, 程永任 |
---|---|
Other Authors: | 李琳山 |
Format: | Others |
Language: | zh-TW |
Published: |
2008
|
Online Access: | http://ndltd.ncl.edu.tw/handle/99061010271406424560 |
Similar Items
-
Minimum Phone Error Training of Acoustic Models and Features for Large Vocabulary Mandarin Speech Recognition
by: Jia-Yu Chen, et al.
Published: (2006) -
An Initial Study on Minimum Phone Error Discriminative Learning of Acoustic Models for Mandarin Large Vocabulary Continuous Speech Recognition
by: Jen-Wei Kuo, et al.
Published: (2005) -
Continuous mandarin speech recognition with very large vocabulary
by: Wang, Xin Min, et al.
Published: (1995) -
Improvement on Language Modeling for Large-vocabulary Mandarin Speech Recognition
by: Chou, Chien-Pang, et al.
Published: (2009) -
Tone Recognition for Fluent Mandarin Speech and Its application on Large Vocabulary Recognition
by: Wan-Yi Lin, et al.
Published: (2004)