Author(s): Mousmita Devi, Prof. P.H.Talukdar
In human speech one of the key aspects is its Formant structure. Formants are resonances of the vocal tract, and as such they have a close relation to the vocal tract geometry. In other words Formant frequency may be refers to the spectral peak of the sound spectrum. Generally, for perception and discrimination three formants namely First (F1), Second (F2) and Third (F3) are considered. In this paper, the formant frequency of Assamese vowels and words of typical structure i.e. CV, CVC, VCV are estimated which can be helpful for developing Assamese Automatic Speech recognition (ASR) system. In this study, we design a database which consist of eight Assamese vowels and words of 5 from each structure total 15 words uttered by 5 times which make our database having total 115 samples. In this study we have analyze the sample for F1, F2 and F3 formant frequencies. It is observed that a significant variation is present in the formant frequencies with respect to gender for both phoneme set as well as the word set.