Articulatory reference theory should play a leading role in the future studies of analysis and recognition of speech. Articulatory representation of speech involves acoustical, anatomical and neurophysiological descriptions. This paper describes two methods of transformation of vocal tract configurations into frequency spectrum as the bases of practical speech analysis on the anatomical level. The first one is the calculation based upon the solution of sound wave propagation through vocal tract with proper boundary conditions this one becomes, however, too complex to solve practically in case of nasal sounds which have a side branch produced by nasal coupling. Another one is the calculation due to the equivalent network analysis of vocal tract configuration and this one gives the values of formant and anti-formant frequencies which are in good agreement with the results of analysis in case of nasal consonants. A few examples of calculation of formant/anti-formant frequencies from given vocal tract configurations in case of vowels, consonants and nasal consonants have confirmed that these two methods provide feasible processes of speech analysis in terms of vocal tract configurations.
抄録全体を表示