In the Brain Computer Interface domain, studies on EEG represent a huge field of interest. Interactive systems that exploit low cost electroencephalographs to control machines are gaining momentum. Such technologies can be useful in the field of music and assisted composition. In this paper, a system that aims to generate four-part polyphonies is proposed. An artificial intelligence algorithm permits to generate polyphonies based on the N. Slonimsky’s theory by elaborating data coming from a Leap Motion device, to detect user’s hand movement, and a five-channel EEG signal detection device.