Segmental feature extraction and coding for speech synthesis

This paper describes a segmental feature extraction and speech coding method in an acousticarticulatory domain using nomograms that represent a mapping between formant frequencies and articulatory parameters. The vocal tract model is a modified Fant model, in which we newly introduced a parameter for successively adjusting vocal tract lengths. We investigated first the relationship between formant contours and those of articulatory parameters and found the effectiveness of the articulatory domain for organizing acoustic-phonetic features with little dependency upon languages. Next, we applied the method to the low bit rate coder and confirmed that good quality speech synthesis was achieved in the condition of 18 bit used for articulatory code words.