Abstract

This is one of a developing series of duos for a human and a machine performer. Both “musicians” adapt to each other through mutual listening (i.e., via audio only) and response as the performance develops. The human’s improvisation is encoded by the computer through statistical analysis of extracted features and by cataloguing these in real time. Each observation made by the computer is assigned to a set of musical output behaviors. Recurring features of the player’s improvisation can then be recognized by the computer. The machine “expresses” this recognition by developing, and modifying, its own musical output, just as another player might.