Music Analysis and Spectral Modeling Based on Cubic B-splines
Daniel Möhlmann, Otthein Herzog, Thomas Wagner · University of Michigan Library Repository · 2009
Structured representations of audio offer interesting opportunities for analysis and manipulation: techniques like spectral modeling help to decompose digital sound recordings into sinusoidal elements, which may then be analyzed or transformed separately.In this paper, we present a novel method to perform high-level decomposition of musical signals to obtain an intuitive and compact structured representation.We present examples of our spline conversion method for both single instrument notes and complex music recordings.Our approach is based on sinusoidal analysis and uses cubic B-splines to encode sound elements in frequency and time.The proposed model preserves the relevant details of the spectral frequency distribution to a large degree and is well suited for additive synthesis.