Score following: An artificially intelligent musical accompanist

Score Following is the process by which a musician can be tracked through their performance

of a piece, for the purpose of accompanying the musician with the appropriate

notes. This tracking is done by following the progress of the musician through the

score (written music) of the piece, using observations of the notes they are playing.

Artificially intelligent musical accompaniment is where a human musician is accompanied

by a computer musician. The computer musician is able to produce musical

accompaniment that relates musically to the human performance.

Hidden Markov Models (HMMs) are a stochastic modelling tool that can be used

to represent real-world systems in a variety of domains.

This project discusses how HMMs can be used in the domain of Score Following

and describes the construction and evaluation of a score following system that

uses HMMs to implement score following. It explores the hypothesis that using an

HMM to represent a musical score is an efficient and practical way to implement score

following, and that in particular this method is suitable for providing real-time accompaniment

to a human performer.

The score followers developed during this project are tested and compared against

other score following systems and against human musicians. The resulting performances

support the project hypothesis to a large extent.