Raphael C., Synthesizing Musical Accompaniments With Bayesian
belief networks, Journal of New Music Research, Vol. 30, No. 1, pp.
59–67, 2001
Abstract
This paper discusses recent work in creating a computer
program that plays the role of a sensitive musical accompanist.
An accompanist must synthesize a number of different
sources of information including a real-time analysis of the
soloist’s acoustic signal, an understanding of the timing relationships
represented in the musical score, the interpretation
of the soloist learned through rehearsals, and musical
constraints on the way in which the accompaniment can be
played. A probabilistic framework is presented in which all
of these knowledge sources can be represented and learned
from actual data. This model then forms the basis of an
approach to musical accompaniment using the machinery of
Bayesian belief networks. A demonstration is provided from
J.S. Bach’s Cantata 12.