Information Technology Reference
In-Depth Information
However, in game audio, this becomes a seduc-
tive goal. Take again the case of the bouncing ball.
Sonically we do not wish to model the nuance of
kinetic, potential and elastic energy exchange, for
entertainment purposes (good enough synthesis)
we are satisfied with a fair approximation at a tim-
ing curve and an adequate model of the modes of a
sphere excited at an exterior point. While the ball
is unseen (acousmatic) no real advantage accrues.
The payoff comes when we have accompanying
graphics provided by a physics engine. Some
of the parameters needed for our synthesis are
available for free, with the added bonus of perfect
synchronisation. So, up to a point, a hard essen-
tialist's goal is an noble one for computer games.
At what point does it become pathological? The
question is whether over-computation is happen-
ing; are we calculating irrelevant information? It's
hard to draw a definite line, but using information
bandwidth studies from HCI as a yardstick, then
once the complexity of the model greatly exceeds
the information that can be acquired by a person
then we're probably going down the wrong road.
difficult tests, the plausibility of the results may
eventually break down. Behaviouralists know up
front that their models will only work for a pre-
defined range of uses. For the bouncing ball, we
might choose to ignore all but a pair of material
parameters and the initial height from which the
ball falls. Given that no other changes occur (the
surface stays fixed, and gravity doesn't change),
then everything else is irrelevant to the subsequent
sound pattern. Constraints must be drawn up
a
priori
and be satisfied throughout the sounding,
which somewhat breaks the conditions of sponta-
neity when we make a simplification that “bakes
in” some future behaviour for a while.
Phenomenal
The school of phenomenology (in the sense of Hus-
serl and Merleau-Ponty) might say: “If it works,
use it”. Meaning, it does not even matter whether
the underlying process matches some counterpart
in reality, or is understood, so long as its facade
yields passably aesthetic results. The senses are
all. Some of Miranda's (2002) [improvements
by Serquera 2010] experiments with cellular au-
tonoma could be seen as phenomenal synthesis
since, although certain generative systems can be
discovered yielding astonishing sonic effects, we
don't have any mapping between their parameters
and sensible, acoustically viable parameters. The
weakness of phenomenal synthesis is that it is
brittle and it works for a small range of sounds or
islands, and it easily breaks when pushed too far.
Its advantage is that if we only pay attention to
the surface features of a sound we can do unusual
synthesis cheaply, as in Serquera's multi-voter
method; a way of herding the CA cats into a pat-
tern that appears to be a natural process.
Yet careful, reasoned use of phenomenal
technique crops up frequently in practical proce-
dural design, and is vital in many places. A chirp
impulse replacing an impossibly loud transient,
the use of noise bursts in critical band masking,
and grain dilation to increase perceived loudness,
behavioural
Behaviouralists are concerned with rendering
convincing facsimiles of sonic phenomena by
understanding the underlying physical behaviour
though not necessarily modelling it directly. The
behaviouralist position admits psychoacoustics
and perceptual science into its framework as a
necessary component. The aims are to produce
perceptual realism (effect, recognition, semantics,
and emotion). The mechanism of the underlying
process is relevant though we are concerned
with data reduction, ignoring all except those
parameters and control systems that are most
important. An analogy may be drawn to Searle's
Chinese Room or to the Turing Test. As a black
box, while we obtain consistent and plausible
results, and while the effects are subjectively
correct, the internal functions are unimportant to
our judgment. As we apply more and increasingly
Search WWH ::
Custom Search