Evaluation of Iterative Methods for Wavetable Matching

Matching musical instrument tones is a classic problem in computer music. Previous work has shown various noniterative methods to be successful. Iterative methods, where the basis spectra are found one by one over several iterations, have received much less attention. Iterative methods are more efficient, with a linear running time. Several iterative methods are evaluated and compared. Matching results for a range of instruments show, surprisingly, that iterative local search can find matches comparable to near-optimal noniterative matches. As an added advantage, iterative local search offers real-time control of the number of wavetables, allowing computation and memory efficiency, and good matching accuracy.