Information Technology Reference
In-Depth Information
Perceptual considerations might, however, be combined with these models to
propose intuitive control strategies as described in the following sections.
4.4.1 Control of Synthesis Parameters
Although physical models can produce high-quality sounds that are useful for
instance for musical purposes, this approach is less adapted to environmental
sounds, when the physics of such sound sources is not suf
ciently well understood
or the existing models are not real-time compatible. In such cases, signal models
that enable the simulation of the sound vibrations through mathematical models are
useful. The control of these models consists in manipulating physical or signal
parameters. Practically, these approaches might involve the control of physical
variables (for instance, characterizing the tribological or mechanical properties of
the source) or a high number of synthesis parameters (up to a hundred) that are
generally not intuitive for a non-expert user. This means that a certain scienti
c (or
musical) expertise is needed to use such models (expert control). In fact, the cali-
bration of the control of these models is based on an
error function
that reveals the
difference between the model and the actual physical sound vibration (cf. Fig.
4.2
).
4.4.2 Control of Perceptual Effects
Common to all the previous approaches described in Sect.
4.4.1
is the lack of
perceptual criteria. Actually, since the timbre of the resulting sound is generally
related to the synthesis parameters in a non-linear way, the control process can
quickly become complicated and non-intuitive. The design of a control of per-
ceptual effects may lead to the de
nition of an intuitive control strategy. In par-
ticular, based on the identi
cation of invariant sound morphologies (cf. Sect.
4.3.2
),
control processes mediating various perceptual evocations could be designed. In
line with the previous de
nitions of structural and transformational invariants, the
framework of our control strategy is based on the so-called {
action
/
object
} para-
digm, assuming that the produced sound can be de
ned as the consequence of an
action on an object. This approach supports the determination of sound morphol-
ogies that carry information about the action and the object, respectively.
Here we present several synthesis tools that we have developed for generating
and intuitively controlling sounds. These synthesis models make it possible to
relevantly resynthesize natural sounds. In practice, we adopted hierarchical levels of
control to route and dispatch the parameters from an intuitive to the algorithmic
level. As these parameters are not independent and might be linked to several signal
properties at a time, the mapping between levels is far from being straightforward.
Sounds from impacted objects
: We have developed an impact sound synthe-
sizer offering an intuitive control strategy based on a three-level architecture
(Aramaki et al.
2010
a) (cf. Fig.
4.1
). The top layer gives the user the possibility to
de
ne the impacted object using verbal descriptions of the object (nature of the
