Evolutionary Music Project
The aim of our research is to compose a human-competitive rhythm accompaniment using interactive evolutionary computation based approach. Evolving of rhythm accompaniment is based on genetic algorithm. Rhythm accompaniment is generated with regard to user specified parameters and with regard to an accompaniment instrument, for instance bass, rhythm piano or rhythm guitar. An innovative representation of rhythm knowledge in genetic algorithm was proposed. Evolution is driven by special fitness operators. They measure quality parameters of generated accompaniment, like rhythm conformity or unconformity with an accompaniment instrument and quantity of rhythm accompaniment. Using crossover and several mutation operators individual rhythm phrases are adapted to satisfy the fitness function.
The GeneticDrummer:
Genetic Drummer is a software application for evolving rhythm accompaniment based on genetic algorithm. It cooperates with Cakewalk Pro Audio or Cakewalk Sonar using MIDI and DDE. In ProAudio or Sonar user interface phrases of an accompaniment instrument are selected and matching rhythm accompaniment is generated by GeneticDrummer and afterwards returned to ProAudio or Sonar as a new track. System generates rhythm grooves or breaks according to user specified parameters. Software is written in Allegro Common LISP.Further details can be found in my master-thesis (in czech):[PostScript file, 2MB] and in several articles, see publications.
GeneticDrummer sample results:
Chicken - [MP3 file, 7MB]A Night In Tunisia - [MP3 file, 3MB]
Blue Bossa - [MP3 file, 3.6MB]
Summertime - [MP3 file, 3MB]
Sockshop - [MP3 file, 3MB]
Following files contains sound representation of the populations used for generating accompaniment in Sockshop song.
Sockshop - population used for groove [MP3 file, 0.6MB]
Sockshop - population used for break [MP3 file, 0.5MB]