Tone recognition for Thai

This paper describes the tone recognition for Thai language. The autocorrelation method is used to extract the pitch or fundamental frequency (FO) from a speech signal. Hidden Markov models (HMMs) are then used to recognize these pitch contours to determine the tones. Though the tone information is superimposed on the vowel portion of a syllable, recognition experiments show that tones and vowels are independent from each other. Tone information extracted from one vowel can be used on another different vowel.