Music Plus One and machine learning

A system for musical accompaniment is pre-sented in which a computer-driven orches-tra follows and learns from a soloist in a concerto-like setting. The system is decom-posed into three modules: the first com-putes a real-time score match using a hid-den Markov model; the second generates the output audio by phase-vocoding a preexisting audio recording; the third provides a link be-tween these two, by predicting future timing evolution using a Kalman filter-like model. Several examples are presented showing the system in action in diverse musical settings. Connections with machine learning are high-lighted, showing current weaknesses and new possible directions. 1. Musical Accompaniment Systems Musical accompaniment systems are computer pro-grams that serve as musical partners for live musi-cians, usually playing a supporting role for music cen-tering around the live player. The types of possi-ble interaction between live player and computer are widely varied. Some approaches create sound by pro-cessing the musician’s audio, often driven by analysis of the audio content itself, perhaps distorting, echo-ing, harmonizing, or commenting on the soloist’s au-dio in largely predefined ways, (Lippe, 2002), (Rowe, 1993). Other orientations are directed toward impro-visatory music, such as jazz, in which the computer follows the outline of a score, perhaps even compos-ing its own musical part “on the fly ” (Dannenberg & Mont-Reynaud, 1987), or evolving as a “call and re-sponse ” in which the computer and human alternate the lead role (Franklin, 2002), (Pachet, 2004). Our focus here is on a third approach that models the tra-ditional “classical ” concerto-type setting in which th