Audio Analysis for Rhythmic Structure
Crawford Tait · 1995
Although wavelet analysis has been shown to be of interest in an audio context, little interpretation of the wavelet coefficients has been attempted. This work utilises a semitone-- based wavelet transform tailored for musical applications, and uses the coefficients to detect note onset locations in monophonic music. Many theories of rhythmic structure are based solely on inter--onset intervals, and the automatic generation of onset times would enable a variety of applications. We describe a method based on pre--processing the modulus of the wavelet detail coefficients, followed by calculation of a vector distance measure. Ways of detecting peaks in this function, and classifying those peaks to identify onsets, are discussed. The results of applying these methods to a variety of musical examples are presented, and it is shown that onsets can be detected even in the presence of overlapping notes. Lastly, the results of an experiment into the effects of envelope and harmonic complexity a...