Abstract:
The purpose of this thesis was to present the Thai consonant pronunciation program for
hearing-impair person by using artificial neural network. Most of hearing-impair person had
unclear pronunciation problem. Their communication could be developed by appropriate,
improving and training. The thesis presented 21 Thai consonant pronunciation program composed
of Thai-speech recognition algorithm which were consisted of finding endpoint detection of
speech to maximum energy of speech by using energy rate with zero crossing rate, feature
extraction using LPC and MFCC, voice energy, delta MFCC and delta Energy. The artificial
neural network was used for Thai consonant recognition and decision in the recognition pattern
step and decision rule testing. The system was trained by 20 males, 20 females prototype and 5
males, 5 females for testing group. The result of the testing group shown accuracy of 80.95% for
males and 74.29% for females.