Odysseus

Computer music is increasingly popular and has the potential to be very expressive. The tools required to create and perform computer music are becoming cheaper all the time. But often creating computer music is a technical challenge rather than an artistic one. Music software is often difficult to use and inaccessible to musicians without computer training. Some musicians respond by relying on pre-programmed beats and sounds, and so fail to explore the full power of the computer as a musical instrument. It is the aim of this project to investigate new, easy-to-use methods of computer music composition and performance.

One idea, common in other fields, is to use "machine learning" methods to help the user. The software tries to "learn" what the user wants to do, and offers some solutions. "Evolutionary computation" or EC is one type of machine learning, based on the idea of biological evolution. The software creates a population of musical objects -- say, drum beats -- and the musician chooses the best ones. The software takes these and breaks them up, then recombines and mutates them to create a new population of beats. The process repeats multiple times, with the beats gradually matching the musician's preferences.

This process is similar to that of a gardener breeding flowers by choosing the best specimens at each generation -- the gardener gets good results but doesn't need to understand the flowers' underlying genetic code. In a similar way, the musician does not have to interact directly with low-level details of the drum beats, but controls the music using a high-level evolutionary interface. The strong analogy between the evolutionary mechanism and the gardener's breeding of flowers makes the interface intuitive and accessible.

This is the approach which will be used here. The musical material will be under the control of an evolutionary interface. The musician's gestures and sounds, captured by a camera and microphone, will be translated into control commands for the music.