Authors : Syed Akhter Hossain , Md Lutfar Rahman and Md Farruk Ahmed
Abstract: Automatic segmentation of Bangla isolated utterances is investigated in this study based on zero crossing and frame energy based segmentation of containing both vowel and consonant phonemes of both male and female speakers. This study proposed an algorithm for automatic segmentation of sections of speech samples based on whether they are silence, voiced or unvoiced speech. The segmentation is accomplished based on the processing of the speech samples and calculation of zero crossing and short-term energy functions. The thresholds are set based on the maximum likely hood for more accurate labeling the parts of speech. The study has shown good accuracy of lassification in the algorithms implemented over all speech samples. Such an effective speech segmentation would play a central role in voice activated application development.
Syed Akhter Hossain , Md Lutfar Rahman and Md Farruk Ahmed, , 2005. Automatic Segmentation of Bangla Speech: A New Approach. Asian Journal of Information Technology, 4: 1127-1130.