Physical Modelling for Sound Synthesis - Game Sound Technology and Player Interaction: Concepts and Developments - page 349

They focused on making realistic object contact
sounds by considering the shape and material of
the sounding object, the location of contact on
the object, and the force of the impact. Sound
is generated by modal synthesis and an object's
modal data is determined by calculation. This is
possible for regularly shaped objects and, in Sonic
Explorer , has been carried out for strings, bars,
plates, and membranes (in physical modelling
terminology, a membrane is a two-dimensional
system with no stiffness). The process of calcu-
lating modal frequencies for these objects is ex-
plained in chapters two and three of The Physics
of Musical Instruments (Fletcher & Rossing, 1991)
and in Sonic Explorer this is carried out off-line
as a pre-processing stage. Doel and Pai explain
that the contribution of each mode to the overall
vibration depends on the impact location and they
derive some general formulas to calculate the mode
amplitudes for a given impact location. This means
that when an object is struck on different locations
there will be subtle differences between the sounds
produced that match what the listener intuitively
expects to hear. The authors do not take into ac-
count the directionality of the emitted sound nor
do they accurately model a physical environment
as this is beyond the scope of their project but,
with a few justified simplifications, they present a
suitable method for transforming the vibrations of
an object into the sound that is heard. Next, they
show how to account for different materials by
explaining that an object's internal friction param-
eter, which is determined by its material, affects
how its sounding frequencies become damped
with time. By changing the damping values they
can therefore create the effect of sounding differ-
ent materials. A frequency-independent damping
value is applied to all frequencies equally and a
frequency-dependent damping has a greater effect
on higher frequencies.
The beginning of the new millennium saw a
significant contribution from a team at the Poly-
technic University of Milan, led by Augusto Sarti.
In their paper Object-Based Sound Synthesis for

Virtual Environments (Pedersini, Sarti, & Tubaro,
2000), they present a way to model multiple in-
teracting sounding objects. They model sounding
objects by combining digital waveguides and wave
digital filters (Fettweis, 1986), which are closely
related to digital waveguides. Nonlinear elements
are incorporated in object contact conditions and
these are exploited to make a dynamic intercon-
nection topology. This allows for connections
between sounding objects to be made and broken
as the acoustic effects are accurately modelled.
They also give an overview of available sound
rendering algorithms, with a strong influence
from 3D graphics rendering. More recently, this
approach has been automated using a binary con-
nection tree (BCT) (De Sanctis, Sarti, Scarparo,
& Tubaro, 2005) allowing real-time interactions
between sounding objects in an interactive setting.
In 2001, Doel and Pai, along with others at
the University of British Columbia, Vancouver,
developed a system for bringing real world objects
into the virtual domain. Scanning Physical Inter-
action Behaviour of 3D Objects (Pai et al., 2001)
extends the idea of simply scanning an object to
create a graphical virtual equivalent by capturing
its deformation behaviour (how it reacts to external
force), its surface texture, and its sound producing
properties. This is achieved by scanning the inter-
action behaviour in a variety of ways using robotic
measurement facilities. While the techniques for
finding an object's deformation behaviour and
surface texture are fascinating they fall outside
the scope of this chapter and so this project will
be examined from an audio perspective.
Similar to The Sounds of Physical Shapes , this
project generates sound by modal synthesis but
now the modal data is automatically determined
for an object based on robotically obtained mea-
surements. To this end, a robotic arm applies an
approximation of an impulse force (a short tap)
to a location on the object being scanned and
the resulting sound is recorded. A technique is
described for extracting the object's modal data
(frequency of vibration modes, the amplitude