Matthias DemoucronDoctorant, Sound synthesis of bowed string instruments
- Tel : 01 44 78 14 86
- Email :
These pages present some material related to my PhD work. You will find sounds, videos and illustrations that could not be included in the paper version. Feedbacks, remarks, suggestions, constructive critics and intellectual supports are welcome...
- PhD thesis: On the control of virtual violins - Physical modeling and control of bowed string instruments
- Measurements of bowing parameters in real performance
- Sound examples: control with real measurements and synthesized gestures
- Control interfaces: real time control with physical interfaces
This work aims at developing some tools for a realistic and "real gesture inspired" sound synthesis of bowed string instruments such as violins. A minimal physical model is developed then implemented as a Max/MSP object for real-time synthesis. Optimal input parameters (the time evolution of bow speed and bow pressure, mainly) are deduced by measurements of the effective control of the musician. In this work, we are focusing on the influence of a realistic control on the realism of the bowed string sounds. Such studies could be performed also with the use of more complex models. However, by using a very basic model we want to demonstrate that the realism of the sounds is hidden just as much in the control of the model as in the model itself.
Physical modelling permits to describe mathematically the functioning of a mechanical system (the musical instrument, in our case). In the specific case of the violin, two "ingredients" have to be taken into account for a minimal explanation of the vibration: the dynamical behaviour of the string (how it reacts to a force), and the interaction force (friction) between the bow and the string. The dynamical behaviour of the string is modeled with a differential equation. The friction force is dependent on the normal force (how much the bow presses on the string) and the relative velocity between the bow and the string. The friction interaction produces an alternation of periods during which either the string "sticks" to the bow-hair, or it slides under the bow-hair.
The physical modelling gives a system of equations that can be computationally solved in order to simulate the motion of the string. The resulting simulation depends on several parameters describing the properties of the mechanical system (string parameters such as length, tension and damping or friction parameters). More interestingly, it can be controlled in real-time by input parameters describing the action of the player, which allows a powerfull interaction with the user.
Measurement of real gestures
Learning how to use the physical model for sound synthesis can be as long as the effective learning of the instrument itself. This part of the work aims at taking advantage of the knowledge of the player in order to get some keys on the use of the model.
For that purpose, measurements are done with musicians performing specific bow strokes or musical extracts, then some features are extracted from the measurements. These features are related to some specific technical points of bowed string instruments playing, to the accuracy of the instrumentalist's control, to the performance of a specific way of playing or to the realisation of some musical ideas.
However, the measurement of real musical performances requires the development of specific devices that are not too disturbing for the player. In the case of violin playing, a specific sensor was built in order to measure the bow force on the string, then combined with a motion capture system.