Automatic Speech Recognition for Vietnamese Using HTK System
Hong Quang Nguyen, Van Loan Trinh, The Dat Le · 2010
This paper presents our research on Vietnamese speech recognition using HTK1 system. Our method is the adaptation of Vietnamese data into HTK system: text corpus was collected from Web and then bigram language model was built by the CMU SLM toolkit. The experiments were carried out with different acoustic models. Further more, the experiment using speaker adaptation technique was implemented by MLLR algorithm. For our results, WA (Word Accuracy) in the case of speaker independent recognition test is 71.37% and 75.96% with the test of speaker adaptation.