Physical Modelling for Sound Synthesis - Game Sound Technology and Player Interaction: Concepts and Developments - page 343

PHYsIcAL MODELLING FOr
sOUND sYNtHEsIs

Procedural audio is generated in real-time
from all the relevant information available. As it
is generated in response to sound producing events
it also fits the term “dynamic audio”, which Col-
lins (2008) defines as audio that “reacts both to
changes in the gameplay environment, and/or ac-
tions taken by the player” (p. 4). Researchers have
developed techniques to generate many types of
sounds from pouring and bubbling water to fire and
thunder and from guns and explosions to animal
and vehicle sounds. Farnell details many of these
techniques in Designing Sound (2008). While it's
true that synthesising sounds from scratch is more
computationally expensive than sample playback
“the rewards are astonishing” (Farnell, 2008, p.
1). The ever increasing capabilities of personal
computers and games consoles coupled with the
desire for realism means that procedural audio
is set to become part of the future of computer
game audio.
Farnell (2011) describes the thinking behind
procedural audio, with the ultimate goal of cre-
ating perceived realism, elsewhere in this topic.
Most often this will involve a degree of sound
synthesis. The current chapter is concerned with a
specific type of sound synthesis known as physical
modelling which is mentioned by Marin D. Wilde
(2004, pg. 158) as a future direction for computer
game audio. To physically model an object means
to simulate how it behaves physically and from
an audio perspective this means simulating how
an object vibrates in response to excitation and
causes sound waves to be radiated from it. Physical
modelling can be used to realise procedural audio
and would be the preferred method of those whom
Farnell (2011) might call “moderate essentialists”.
A range of physical modelling techniques have
been developed and this is the subject of the next
section.

Sound synthesis techniques have been developed
for musical and sound design purposes since the
late 1950s when Matthews first performed wa-
vetable synthesis by generating sound from data
stored in tables (Bilbao, 2009). The following
decade saw the development of FM synthesis and
additive synthesis and, since then, composers and
sound designers have learned to use these tech-
niques and others to achieve their aims. However,
these techniques, often described as abstract, have
no basis in the real world and so the link between
the input parameters and the sound produced is
not naturally intuitive. This means control systems
are often complex for the user (Adrien, 1991) and,
while commonly used in early computer games,
there was no obvious link between their input
parameters and in-game variables. Physical mod-
elling synthesis techniques are, however, based
on real physical systems where there will usually
be an intuitive link between the configuration of
the system and the sound produced. Hence there
should be an intuitive link between the parameters
of a physical model and the sounds it produces,
offering good control to sound designers (and
musical composers).
To discuss physical modelling from an acoustic
point of view is to discuss the physical modelling
of musical instruments as this is a long established
research area in which much progress has been
made. Much of what has been learned can be ap-
plied to the physical modelling of any sounding
object and is therefore relevant where sound effects
are to be created in a virtual environment. This
section looks at some of the methods available
before giving a more detailed look at the methods
most suitable for operating in real-time and those
used in the projects discussed.
Throughout history, musical instruments have
evolved to create sound in a sometimes complex
way while giving relatively simple control to the
musician. Instrument makers have, through cen-