Mahidol University Logo
Faculty of ICT, Mahidol University
 

Admissions

Printable Version

 

A PROTOTYPE OF SPEECH RECOGNITION SYSTEM FOR THAI LANGUAGE WITH SPEAKER INDEPENDENCE

 

TITLE A PROTOTYPE OF SPEECH RECOGNITION SYSTEM FOR THAI LANGUAGE WITH SPEAKER INDEPENDENCE
AUTHOR THANYARAT PRUTPAPOP
DEGREE MASTER OF SCIENCE PROGRAMME IN COMPUTER SCIENCE
FACULTY FACULTY OF SCIENCE
ADVISOR SUPACHAI TANGWONGSAN
CO-ADVISOR SUKANYA PHONGSUPHAP
CHOMTIP PORNPANOMCHAI
 
ABSTRACT
This research presents the development of a prototype of speech recognition for the Thai language with the class of speaker independence. The main concept is based on the Hidden Markov Model (HMM) on the acoustic level. The system makes use of a tool known as Hidden Markov toolkit, which is the instrument employed to generate the model in both steps of training and recognition. However, as the tool was originally designed for English language, in order to make it applicable, Thai speech syllables had to be first mapped into standard codes of International Phonetic Alphabet (IPA) accordingly. The next steps were the application of syntactic and semantic rules in the prototype in order to improve the accuracy rate of recognition. In the experiment, 20 people, 10 males and 10 females were invited, and a set of 128 simple vocabulary words commonly formed at the Primary 4 level were used for training. Then 6 new people were invited to conduct the testing by speaking several sets of conversation in meaningful sentences based on those 128 vocabularies. The results showed that on the acoustic level, only 55.21% (Top 1) and 57.39% (Top 3) are obtained with no surprise. However, after applying the syntactic rules for further processing, the result shows a better improvement as 66.52% (Top 1) and 73.04% (Top 3) were obtained. Finally, with the semantic rules for filtration in the process, the results showed a significant improvement as 75.65% (Top 1) and 86.52% (Top 3) were obtained respectively. In conclusion, the developed prototype has met the set objectives accordingly and satisfactorily.
KEYWORD SPEECH RECOGNITION/ HIDDEN MARKOV MODEL

 

Go to Top

 

ICT Building, Mahidol University, 999 Phuttamonthon 4 Road, Salaya, Nakhonpathom 73170 Tel. +66 02 441-0909 Fax. +66 02 849-6099
Mahidol University Computing Center, The Faculty of ICT, Mahidol University , Rama 6 Road, Rajathevi, Bangkok 10400 Tel. +66 02 354-4333 Fax. +66 02 354-7333