Interpreting Rhythmic Structures Using Artificial Neural Networks
Ph.D. Thesis by
S. C. Roberts
(formerly at)
Department
of Physics and Astronomy
University
of Wales, College of Cardiff
September 1996
Introduction
An artificial neural system for the real-time interpretation of rhythmic
structures is developed. The self-organising system responds to a
continuous sequence of musical events and discovers patterns of event onsets,
in a manner which reflects temporal grouping in rhythm perception.
Patterns are encoded by the SONNET network devised by Albert Nigrin.
A neural foot-tapper enables patterns to be encoded from expressive
performances. Software simulations of the system are discussed throughout
the thesis and SONNET modifications are presented.
Download
The thesis is available as compressed PostScript. Don't be put off
when I say it is 333 pages long - it includes many figures, 25 pages of
appendices and a 21 page bibliography.
To download:
Click the icon
below and save the linked file as irsuann.ps.Z
Uncompress to
give irsuann.ps with uncompress irsuann.ps.Z (UNIX) or
WinZip (PC).
irsuann.ps
is ready for printing (you may need Ghostscript
to view PostScript).
irsuann.ps.Z (908K)
Links
SONNET
Rhythm perception
You are visitor number 
.
(Or maybe you have visited this many times.)