Evaluation Of Multiresolution Representations Of Musical Rhythm
A dynamic representation of musical rhythm, the multiresolution analysis using the continuous wavelet transform (CWT), is evaluated using a dataset of the interonset intervals of 105 national anthem rhythms. This representation decomposes the temporal structure of a musical rhythm into time varying frequency components in the rhythmic frequency range (sample rate of 200Hz). Evidence is presented that the beat (typically quarter-note or crochet) and the bar (measure) durations of each rhythm are revealed by this transform. Such evidence suggests that the pattern of time intervals, when analyzed with the CWT, function as features that are used in the process of forming a metrical interpretation. Since the CWT is an invertible transform of the interonset intervals in each rhythm, this result is interpreted as setting a minimum capability of discrimination that any perceptual model of beat or meter can achieve. It indicates that a bottom-up, data-oriented process (or a non-cognitive model) is able to reveal durations which match metrical structure from realistic musical examples. This then characterises the data and behaviour of a top-down cognitive model which must interact with the bottom-up process.
Leigh M. Smith and Henkjan Honing
Proceedings of the International Conference on Music Communication Science, 4 pages, Sydney, 2007