|
The goal of this project is to develop an elegant, general and unifying strategy for a block-wise design of nonlinear physical and pseudo-physical models in sound synthesis.
The strategy is based on a "divide-and-conquer" approach, in which the elements of the structure are individually modelled and discretised, while their interaction topology is separately implemented in a dynamical and "physically sound" fashion. This novel methodology is used for the development of a new generation of object-oriented physical/pseudo-physical sound synthesis solutions.
The project focused on two application scenarios :
- Virtual musical instruments : object-based acoustic interactions between functional elements in musical acoustics.
- Physical modelling of sounds in virtual reality : multi-modal object-based virtual acoustics
The solutions have been developed in such a way to satisfy a number of strong requirements:
- Physicality : the behaviour of the model meets the expectations and behave in an inherently consistent and stable fashion.
- Flexibility :
- object-based : objects are modelled individually in order to make the approach suitable for object-based virtual environments and in order to be able to use models that are already available in the literature.
- hierarchical : complex models are described by the interaction between macro-blocks, which in turn can be exploded in the interaction between smaller sub-blocks. This explosion process may continue until we reach the highest level of resolution (the lowest logical level), made of elementary blocks such as masses, springs, frictions, acoustic tubes, etc.
- semantic : the global description of the individual building blocks and of their interaction is given in symbolic form, using a meta-language developed for this purpose; this enables the adoption of a Graphical User Interface.
- dynamical : the models are time-varying in all respects :
- time-varying modelling parameters
- time-varying topology of interaction (individual models can freely interact with each others)
- open : the system enables the interaction between physical or pseudo-physical blocks (described by pairs of dual variables)
- Usability : the object interaction is modelled and controlled through a minimum number of significant user-oriented control parameters. The system also enables dynamically changing topologies of time-varying interactions (on-the-fly fusion of independent systems through the monitoring of “proximity” and “contact” conditions between objects).
- Intrinsic robustness (“Physically sound” physical sound modelling) : the discretisation process preserves the stability properties of the reference analogic sound generation model. This means that we can choose the temporal sampling rate only with reference to the (vibrational) signals that are being exchanged between blocks regardless of discretisation issues.
- Computational efficiency : the real-time modelling of acoustic interactions between objects requires simple, effective and parallelisable algorithms. Working with an object-based approach is a natural way of creating a parallelisable algorithmic structure. A live performance system is based on 4 PCs (P4-2.8GHz) running four synthesis engines in real time. This is sufficient for performing live with 10-15 high-quality musical instruments.
|
