Thai speech recognition using Neuro-fuzzy system

There are many popular algorithms to recognize the human voice. The good algorithm not only results the high recognition accuracy, but also robust to noises. Several experiments are done in this research to verify the performance of the Neuro-fuzzy system to recognize the human voice. Eight words in Thai language recorded in a different environment, syllable and pronunciations are used as a data set in the experiments. All words are in machine command category, the meaning of words such as forward, back, left, right and so on, with the purpose of applying to some instrument for disabled people in the future. The preliminary results show that each factor has different effects to the recognition accuracy. However the experimental results show that the Neuro-fuzzy system is quite robust to noise and can yield the higher recognition results compare with other popular algorithms. The methodology of speech recognition proposed in this paper can be one of the best alternative choices to apply to a relative work in the future.