Score following and alignment

Score following and score alignment enables precise moments to be determined within an audio flux, along with the values of associated events marked in the score (length of note/chord attack, end of notes, exact pitch, timbre etc). While score alignment is executed independently from time using the entire audio recording and the entire score, score following, on the other hand, is carried out in real time as the audio flux is playing.

Alignment can be used in synthesis (to encode the nuances of a performance very precisely, for example in a MIDI file extended to supply a high quality synthesis), in musicology (to study different performances of a same work), in education (to show students a visual representation of their execution), in indexing (to constitute learning databases), etc.

For performances of contemporary works that combine instrumental parts with electronics, the technique of score following is often used to synchronize the electronic accompaniment with the instrumental soloist. The score is recorded onto a computer in a specific format containing both the essential elements of the original score and the new information concerning the tracking of the accompaniment. During the piece's performance, real-time analysis of the sound and/or of the gesture information captured from the performer is compared with the recorded score. For every moment of the performance, the score following algorithm determines the corresponding moment in the score and synchronizes the processes programmed in the works electronic part.

Since 1986 IRCAM has been developing score following techniques capable of working with various different instruments including the flute, clarinet, voice and more recently the violin for compositions by Pierre Boulez and Philippe Manoury.

New Max/MSP objects for score following are now available. Grounded in new methods using Hidden Markov Models, these objects are more robust and enable previously non-accessible events (trills, chords) to be synchronized. Projects using this technology, recently developed by the team, are already underway. Development of this technology targets its adaptation to different instrument families and an expansion of its capacity to recognize musical forms not limited to note successions or to a tempo.

Participants