This Master’s thesis work presents an algorithmic approach for automatic synthesis
of the violin’s instrumental gesture controls. Focusing on
” Détaché”,
as a sustained bow motion technique, this work aims to give an instrumental
gesture model generator (a physical performer model) to learn an appropriate
representative function based on a group of already available contours that
belong to different performances of the same bowing technique. Contours are
main physical parameters of bowing which can be used as inputs of a physical
sound model, i.e virtual violin, in order to produce the expected sound. During
this work the primary physical parameters are under study, In particular
bow
transversal velocity
(v), bow pressing force(f) and bow-bridge distance(d).
The result of this work then can be generalized to cover the other bowing
techniques, in particular sustained techniques like different derivatives of
legato. B-splines were used in order to devise a parametric, flexible and robust
model which is capable in capturing subtle variations in the time continuum
of input parameters. A series of test experiments has been conducted in order
to evaluate the learned model, where people with strong background in music
and a considerable proficiency in playing musical instruments were asked to
evaluate the sounds produced by the synthetic model.
The results of evaluation were to high extent satisfactory, indicating that
the quality of the sound produced as the result of artificial gestures (outputs of
performer model) were close enough to the sound produced by the gestures of
experienced violin players.
i