Computational Challenges in Multiple Wavetable Interpolation Synthesis

摘要

A new music analysis/synthesis algorithm, optimized multiple wavetable interpolation, poses significant computational challenges in the spectral matching stage, in which it searches for the subset of the available wavetables that best matches the spectrum at each breakpoint of a piecewise linear approximation of the spectral envelope of a recorded tone. Two methods of reducing the computational cost of spectral matching are presented: a multi-level pruned search and a caching genetic algorithm.