Information Technology Reference
In-Depth Information
There are several fundamental wirings of the three modules, with or without a
human controller (Fig.
6.1
), that can be used to form a taxonomy of computer music
systems. The figure shows the controller placed to the left of the system (parameter
domain) and the audio environment,
Ψ
, to the right of the system. Musicians, oper-
ating in the sonic domain (to the right of the system in the figure) contribute directly
to
Ψ
.
P
and
Q
are established subcomponents of music systems. The novel aspect of
a Live Algorithm derives from the inclusion of a patterning/reasoning module,
f
,
which has neither audio input or output, but is a more general purpose algorithm
which could be applied equally in non-computer music contexts. In general
f
em-
bodies a computational process with input and output parameter streams. In Live
Algorithm terms,
f
is a generative unit, the machine equivalent of ideas and imag-
ination. This function is key to enabling the system to demonstrate the capabilities
of autonomy and novelty.
Each wiring is already in common use in various computer music scenarios.
These are described in the following in each case, and their potential for Live Algo-
rithms research is discussed.
P
performs analysis of incoming audio (Fig.
6.1
A). Its human-equivalent func-
tion is
listening
. In the figure,
Ψ
is the musical environment;
Ψ
in
(
Ψ
out
)arein-
coming (outgoing) audio streams. (Alternatively, an incoming sound wave could be
digitised by an analogue-to-digital converter. Such a converter would be regarded
as part of
P
itself.)
P
processes incoming samples, producing analysis
parameters
.
These parameters seek to describe the audio, in music theoretic terms (events, pitch,
duration), as spectral data, in timbral terms such as smoothness and roughness, or
in other high level descriptors.
P
therefore emits a stream of parameters at a slower
rate than the signal rate. In
Music Information Retrieval
, the data is used for the
automatic scoring of performance. Figure
6.1
A could represent a possible perfor-
mance scenario in which a musician can inspect analysis parameters in real-time,
most likely via a graphic display. This set-up may be used to supplement the sonic
information the musician already has access to. Figuratively, the musician (function-
ing as a controller) is placed to the left of the
P
module to emphasise that system
interaction is via parameters, and not by audio (in which case the musician would
be placed to the right of
Ψ
). Reliable algorithms for machine listening are of con-
siderable importance but the problem is very challenging when the audio source is
a combination of several instruments. Machine listening is the subject of a large
research effort within the DSP community.
P
itself does not perform any function other than analysis. If some further pur-
pose is intended, the analysis parameters are fed into an algorithmic module,
f
,as
depicted in Fig.
6.1
B. For example, if the information is to be used to access similar
excerpts from a music database,
f
would perform the similarity measure and the
look-up.
Note that links between modules are not directional to indicate that parameters
might be passed in either direction. For example, a subcomponent of
f
might require
a finer level of analysis from
P
, and could therefore send an instruction to
P
to that
effect. The bi-directionality of system components means that the division into
P
,
