Information Technology Reference
In-Depth Information
can't just play procedural audio; you have to
run
it,
in a context that makes sense.
other examinations of design, this continues in a
“dialectic” testing the new synthesis against the
internal model and external reference, and iterating
over successive improvements. There are several
flavours of analysis that can be employed, such
as perceptual analysis and signal (correlation,
spectrum, transient) analysis. The one we will
consider in this chapter is physical component
analysis. Without an interpretive intermediate
stage, analysis-synthesis is merely transcoding,
in the sense of a Shannon-Weaver signal. In other
words we could resynthesise a sound transpar-
ently from an analysis and gain nothing (except
data reduction useful for transmission). A proper
analysis yields behavioural features and concepts
that can be meaningfully manipulated to give new
sound transformations. We get proper handles on
things. An analysis that makes sense will allow us
to creatively intervene in a reasoned way because
we are able to abstract something useful from
the example sounds. This is dealt with in several
places in
Representations of Musical Signals
(De
Poli et al 1991).
An analysis works best when carried out over
a long time, or with numerous examples of the
target rather than with a single snapshot. Like
tomography, it is an attempt to reveal what is
hidden within by integrating views. Imagine a
wrapped present that you shake, weigh and tap,
trying to guess what is inside. A stimulus provokes
a response. Consistent responses, mapping domain
to range, constitute a behaviour and, as we think
of the object as a “black box”, by observing the
common responses to lots of environmental stimuli
we thus reveal something about the object. In the
case of an idiophonic object, say a snare drum, the
response may be immediate and consistent, while
the increasing complexity of machines and living
organisms may require deeper analysis and yield
less predictable behaviour. At the limits of behav-
iourism, we can either open the box to reverse
engineer the system or use signal (superficial)
analysis with a phenomenal synthetic method.
There may also be accidental characteristics,
Procedural Audio is More
than Physical Modelling
Since it is often conflated precisely with the words
synthesis or physical modelling, let's be clear
that procedural audio isn't necessarily either of
these subjects. Physical modelling (insomuch as
it refers to finite elements, tensor matrices, mass
spring damper systems, waveguides and other
implementations) and synthesis (as it pertains
to known parametric methods such as additive,
granular or non-linear), are pieces of a larger
picture. This picture also comprises psychology
(perceptual psychoacoustics and auditory cogni-
tion), philosophy and epistemology, and object
domain knowledge.
It is in their combination that we attempt to
reduce a sound to its behavioural realisation. The
culmination of these disciplines is to program
physically informed
synthetic or composite source
sound objects with
behaviourally informed
re-
sponses to input conditions. This involves some
measure of simplification and necessary under-
standing and is not an attempt to construct a one-
to-one model of a system. So, by programming,
I do not mean only models, synthesis methods,
and implementations but, rather, the total set of
principles required to get good behaviour, effective
over a range of use cases. In this respect, procedural
audio intersects with some of the interesting stuff
of orthodox of sound design, the tricks, shortcuts,
perceptual devices, simplifications, and decep-
tions of psychology and storytelling.
bEHAVIOUr AND ANALYsIs
Starting with a concept, analysis is the beginning of
the real work. It seeds mental prototypical designs
which in turn lead to synthesis. As described in
Designing Sound, Computer Sound Design
and
Search WWH ::
Custom Search