鼻子音の分離について

概要

論文の詳細を見る
In words and sentences, the effects of context make the problems about synthesis and recognition of speech more complicated. Then we must inquire into quantitative relations about them. In this paper, from the data of the word consisting of two syllables of V_1NV_2 (V : vowel, N : nasal consonant), we discussed the influence of the preceding vowel upon the second formant frequency (F_<2l>) at the beginning of the following vowel which play an important role in the recognition and synthesis of nasals and tried to represent this second formant by the experimental formula F_<2l>=κ(F_<2i>-F_<2t>)e^<-βτ>+F_<2t> where F_<2i> [cps] is the second formant frequency of V_1 at the boundary between V_1 and N; F_<2t> [cps]means the ideal target frequency of F_<2l> which indicates nearly F_<2l> of the monosyllable; τ is a nasal-segment duration in milli-seconds; κ, β are constants. To the contrary, F_<2t> in the above formula is calculated by F_<2i>, F_<2l> and τ extracted from the sonagram or the autocorrelation function. And this F_<2t> makes possible the discrimination among /m/, /n/ and /η/ for each following vowel.
山梨大学の論文