Sound Synthesis by Simulation of Physical Models of Musical Instruments

Over the last decades, several powerful sound sythesis techniques have appeared on the consumer market. This is due to both the increased computation power of processors and the re-emergence of software-based synthesis. At the same time, the need for simple user interfaces to control complex intrument models has become apparent. In this thesis, the digital waveguide technique is used to build models for wind and string instruments. These models allow the simple use of natural haptic interfaces for the measurement of the input parameters, without the need of complex parameter mapping. This thesis consists of three parts. In the first part, a general introduction is given on musical acoustics and the digital waveguide technique for the discretetime simulation of musical instruments. In the second part, the focus is on different modeling techniques for wind instruments. Starting from an acoustic model, a new digital waveguide model class with formant filtering is constructed, and several multi-note structures are studied. An implementation using the MPEG-4 SAOL is then discussed. The third part deals with the string instruments. First, a single-string model and its associated parameter estimation routines are described. A new theoretical framework for coupled digital waveguide systems is designed and applied to the guitar. This results in a fully coupled instrument model. Finally, a three-layer SAOL implementation is presented, including a player model, a score interpreter, and a core signal processing routine.