Spectral Matching of Musical Instrument Tones

Abstract

This thesis shows how to solve two classic problems in computer music: finding parameters for wavetable and frequency modulation (FM) synthesis that reproduce a particular tone. Previous FM and wavetable matching methods have used trial-and-error or computationally expensive techniques in matching sounds.

This work introduces efficient genetic algorithm-based techniques for determining parameters for the synthesis models. It also describes a principal components-based wavetable matching method and compares it with the genetic algorithm-based technique. Finally, it shows how to generalize these techniques to match a set of musical instrument tones.