Under Construction @ Keele 2016 Volume 1 Issue 1 | Page 52

42
!

reproduction. In the case of one parent reproduction, the variation is created through
a number of bit flips in the binary string. This procedure is known as mutation. In the
case of two parent reproduction, random cut points are selected for the two parent
genotypes and they are then spliced together. This procedure is known as crossover
(Figure 1).2

Initial Random Population – A number of the sounds are
likely to be unusable, a high mutation rate is used to
explore the search space.
The mutation rate is reduced as sounds with some
suitable qualities are discovered.
At this point, sounds with suitable qualities will have
been discovered. Crossover is used to preserve them. A
lower mutation rate is used to explore the nearby search
space.
A suitable sound is discovered, a low mutation rate is
used to generate variations on this sound.
Several sounds from this generation will act as variations
to the parent sound (this assumes the material is being
used for the purposes of composition). This process can
be repeated if further variations are required.

Figure 1: The operation of an interactive genetic algorithm.

2.2. Existing Systems
It is impossible to list all systems that make use of genetic algorithms to control
sound synthesis and processing techniques, so this paper will focus on those that
have informed the development of ESDE.
Colin Johnson makes use of an interactive genetic algorithm to construct an
interface for the CSound FOF granular synthesis algorithm (audio programming
language). This system applies a domain specific approach.3 The user assigns a
numerical rating to each sound in a generation, the higher ranking sounds have a
higher probability of reproducing. This system demonstrates an effective means of
mapping binary strings onto synthesis parameters and allows the user to evolve
usable sounds in relatively few generations. It also provides an example of an
!!!!!!!!!!!!!!!!!!!!!!!!!!!!! HHHHHHHHHHHHHHHHHHHHHHHHHHHHHHHB