Automatic Structure Analysis
When we listen to music, we often notice relationships such as repetition, transposition, call and response, rhythmic patterns, melodic motives, and harmonic sequences. All of these are examples of music structure. This research aims to discover structure within music, using "real" audio signals (as opposed to music notation or MIDI data) as input.
To date, we have used computers to analyse a variety of musical examples. Our analysis works by spotting repeating sequences of music. This in turn is accomplished by analyzing music to obtain features that generally contain information about the harmony and melody of the music over a 25 to 250ms window.
Future work will attempt to utilize information on "musical texture" to improve the analysis. Also, we plan to look for more music relationships, including transposition and rhythmic patterns.