Models of Speech Production for Speech Analysis and Synthesis

Coding (i.e., analysis and re-synthesis) of speech signals on the basis of physiological models of the speech production mechanisms has received considerable attention during the past several years. One of the reasons for this renewed interest is that the known coding algorithms gives unacceptably poor quality of speech at low bit rates (e.g., at 2400 bits/sec). A coding scheme which mimics human speech production may have advantages at such low bit rate. Basically three aspects of speech production need to be modeled. First, the geometry of the vocal and nasal tracts needs to be parametrized. Second, a model must be selected to describe wave propagation in the tract.