One of the basic design considerations for the ASR(Automatic Speech Recognition) is to decide a recognition unit such as phoneme, syllable, word.
In this thesis, the efficient syllable segmentation method of the CVC form of the korean syllable is proposed. The syllable segmentation consists of localization of the vowel segment and decision of the syllable boundary. The vowel segment is detected by classifying the speech signal into voiced and unvoiced sound and by classifying the voiced sound into vowel and voiced consonant segment. Thus, syllable boundary is located somewhere between the two vowel segment, according to the characteristics of the final sounds such as stop consonant and voiced consonant. Finally the simulation result of the segmentation and recognition of the syllable to be segmented are discussed.
