Abstract:
Efficient model-based synthesis of string instruments is typically based on digital waveguides or extensions of the Karplus--Strong (KS) model. The detailed relation of these two cases is analyzed in this presentation by deriving the KS model from a more physical waveguide formulation. The equivalence of these models is discussed and their computational efficiencies are compared. For most instruments and playing conditions KS-type models are preferable from the point of view of fast real-time implementations. In some cases the waveguide model is necessary for detailed synthesis, such as for instruments exhibiting nonlinearities, since it is capable of properly simulating true wave propagation. Methods to extend the KS model in order to include beats and some other nontrivial behaviors are discussed. Finally, the capacity of present signal and RISC processors to implement real-time sound synthesis is estimated. [Work supported by the Academy of Finland.]