Score Following History

From IMTR

This page documents the evolution of Score Following at Ircam since 1983 and in chronological order:


Contents

Early Days

Score Following research was initiated at Ircam by Barry Vercoe in 1983, and also by Roger Dannenberg at CMU. Due to computing limitations at the time, the technology was limited to symbolic inputs from the musician rather than audio. In the case of Vercoe's synthetic performer, Larry Beauregard (the flutist of the Ensemble Intercontemporain) setup sensors on his instruments at IRCAM to provide symbolic inputs to the machine. Vercoe and Puckette extended their synthetic performer by making it train itself to a performance.

MIDI Period

With the advent of MIDI standard in the 1980s and its integration within commercial musical instruments, score followers were adapted to accept symbolic inputs in MIDI format. Many early musical examples of score following used the MIDI version. A historical musical example of this development is Philippe Manoury's Pluton for Piano and Live electronics. Today we tend to use polyphonic score followers directly using live audio for such setups.


Pitch Period

With the advent of dedicated audio processing hardware modules (such as 4X and ISPW at Ircam), the first pitch-based score followers emerged out, this time taking audio as input and following pitches in the music score. Many early real-time electronic pieces used this technology for performances of live electronics with musicians playing a music score. A historical example of this development is Philippe Manoury's Jupiter for flute and live electronics composed originally for MIDI flute and ported for audio in 1992, and En Echo for voice and live electronics. Jupiter is probably the first realtime piece composed in MaxMSP.

HMM Period

Late 1990s saw the advent of probabilistic methods for speech and audio processing. An ideal score follower should take into account uncertainties due to performance or machine perception, thus favoring probabilistic methods for robustness. This led to a new generation of score followers based on Hidden Markov Models (HMMs), started in 2000 at Ircam by Nicola Orio and Diemo Schwarz and led to suivi~ module, enhanced with an artificial training system. This system was first employed in a concert situation in 2005 for a performance of Pierre Boulez' piece ...explosante fixe... for flute, orchestra and electronics.

These developments led to wider use and consideration of score following techniques by composers willing to extend this paradigm to more sophisticated performance and live electronics and paved the way for the latest score following technology using Anticipatory techniques.

Anticipatory Score Following

The musical and scientific goals of score following are different. The scientific one requires exact alignment whereas the musical one demands access to live interpretation parameters to undertake electronic actions synchronous to live performers. In late 2007, in collaboration with composer Marco Stroppa, score following moved to an anticipatory paradigm, decoding both position and tempo in realtime; and anticipating performance parameters as musicians do. These considerations led to the development of Antescofo, which has become the standard score following platform in many pieces involving live electronics.

Synchronous Programming

Given the robustness of Antescofo live detection, the musical goals of such systems became more and more explicit. This led to the extension of score following paradigm to Synchronous Programming. Electronic events are now polyphonic programs, running concurrently and in parallel to the performer, and written in relative time. The attempt is thus to bridge the gap between the performative and compositional aspects of computer music. Antescofo's synchronous programming language has since evolved to address more composers' demands.
Personal tools