Artificial Music Composer
Juan Jesus Romero Cardalda J.J.
Abstract
Traditional Musical Computation Systems had to face the differences
between the computational techniques and the characteristics of musical
creation. Characteristics such as a high degree of subjectivity, a great
irrational component, and a learning process based on the use of examples
and environmental absorption, have made music difficult to be formalized
through algorithmic methods or classical Artificial Intelligence methods
such as Expert Systems.
We propose the creation of a cybernetic model of a human composer in a
primeval stage of human musical evolution, following a paradigm of
cognitive complex models creation, based on the use of the human
reference, not only in a static point of view but also considering its
evolution through time. Therefore, the proposed system simulates musical
creation in one of the first stages of musical evolution, whose main
characteristics are the percussive and choral aspects.
The system is based on Genetic Algorithms, whose genetic population is
integrated by several tribes. This model carries out the task of musical
composition, led by the user who expresses his/her musical taste
assigning a punctuation to each tribe.
The GA selects the worse tribes as individuals to be eliminated. In order
to select those tribes which are going to be used as parents, a random
function is used, having each tribe a probality proportional to its
punctuation. The new tribe is produced by crossing the parent tribes in
each individual. Afterwards, mutation takes place in the created
individuals.
The experiments carried out with this system have proved its
functionality in the composition of rhythmic patterns. It is intended to
enlarge the experiment's scope by communicating the system via Internet.
This would enable its use by users of different musical cultures, taking
into account that the system is user-friendly, since it requires no
musical knowledge.
RNASA LAb. email: jj@udc.es
(Laboratory for Artificial Neural email Tlfno:919855417@activajoven.tsm.es
Networks and Adaptative Systems.) http://rnasa.dc.fi.udc.es/cm
Department of Computing.
School of Computer Science.
University of A Coruņa.
Spain.