5th International Conference on Spoken Language Processing

Sydney, Australia
November 30 - December 4, 1998

A New Strategy of Fuzzy-Neural Network for Thai Numeral Speech Recognition

Chai Wutiwiwatchai (1), Somchai Jitapunkul (1), Visarut Ahkuputra (1), Ekkarit Maneenoi (1), Sudaporn Luksaneeyanawin (2)

(1) Department of Electrical Engineering, Chulalongkorn University, Thailand
(2) Department of Linguistics, Chulalongkorn University, Thailand

In this research, a new strategy of Fuzzy-Neural Network system was proposed for Thai numeral speech recognition. Instead of using the fuzzy membership input with class membership desired-output during training procedure as proposed by several researches, we used the fuzzy membership input with fundamental binary desired-output. This can reduce the misunderstood training, decrease the training time and also improve the recognition ability. The system was tested on the Thai ten-numeral speech (0-9) recognition. The error rate for speaker-independent test achieved 9.2% compared to 14% error rate of conventional neural network system while the error rate of the system using class membership desired-output is quite high because of misunderstood training.

Full Paper

Bibliographic reference.  Wutiwiwatchai, Chai / Jitapunkul, Somchai / Ahkuputra, Visarut / Maneenoi, Ekkarit / Luksaneeyanawin, Sudaporn (1998): "A new strategy of fuzzy-neural network for Thai numeral speech recognition", In ICSLP-1998, paper 0349.