Music database home
The tunes
The PDF pages
[The scores]
[Browse the tunes]
[Find similar tunes]
[The statistics]
[The database]
Storing music
Music analysis

Options in [] I'm working on.

The analysis we can do includes the following.

Statistical analysis

• Analysis by note length distribution.
• Analysis by note pitch distribution (relative to the current key).
• Analysis by the distribution of the intervals between successive notes.
• Analysis by the distribution of successive pairs of intervals between notes.

Results from the above analyses can then be combined with standard cluster analysis techniques or multi-dimensional scaling to determine the ethnic similarities between tunes in our database.

Music harmonisation
We also have programs to generate the type of chord harmonies used in the database of British folk music. Alternative harmonies are optimised using a combination of abstract theoretical techniques (e.g. matching the harmonics of the notes in a phrase, with appropriate weighting functions, to those of the notes in each possible chord) combined with heuristics (e.g. the acceptability of certain chords w.r.t the present key, or the acceptability of certain resolutions). Techniques such as critical path analysis are used in the final optimisation. (See The Harmonisation of Melodies by Computer, Eric Foxley, Proc Second Symposium International Informatique et Musicologie, 1982, Paris).

The harmonies produced can be used in various ways.

• As a basis for harmonies to be used when playing the music. New non-standard harmonies often appear, which are perfectly acceptable.
• As the basis of further clustering, to determine which tunes are possible derivatives of others.
• As the basis for awarding marks for given harmonies. This is the basis of an experimental music harmony course being built in the Ceilidh or CourseMarker system at the University of Nottingham Computer Science Department.