Student project: Analysing Bach's counterpoint with Machine Learning

Suitable For

M.Sc. Conversion (Research Project)
M.Sc. Global
4th year final project
3d year final project

Number of People

Synopsis

Music is a rich dynamic domain with many layers of abstraaction. It can be used as a testbed for data mining techniques for time-dependent data. At the same time, it is an interesting domain in itself for applying data mining techniques.

Bach's music makes heavy use of counterpoint (the use of several independent melodies in parallel). Typically, counterpoint is developed from a single theme, as in canons and fugues, which occurs repeatedly in all parts, possibly undergoing transformations (transposition, inversion, stretto, etc.). A first goal for this project is to develop a system that can recognise the theme and all its occurrences in a piece of music. A more ambitious long-term goal might be to use similar techniques to analyse the harmonic structure of the piece (cadence points, modulations, ect.).

Objectives

To explore the use of machine learning techniques in musical analysis;
to refine existing techniques for mining time-dependent data.

Deliverables

An operational system that can read MIDI files and apply data mining techniques to reveal counterpoint.

Skills Required

competence in music (MIDI, notation, counterpoint)
a general understanding of artificial intelligence and machine learning

Tools/Environments

still to be decided

Benefit to Student

To take part in some genuine and valuable artificial intelligence research
Being able to employ musical skills in a computer science project

Contact

Peter.Flach@cs.bris.ac.uk

Background

http://www.cs.bris.ac.uk/Research/MachineLearning/
http://www.cs.helsinki.fi/~mannila/data-mining-publications.html#sequences