The Impact of Generated and Expressive Modulation of the Synthetic Instrument Sound Parameters on the Impression of Naturalness

Abstract

Despite their different spectral structures, the sound of early instruments from the electrophone group was often considered to be deceptively similar to the sound of wind or bowed string instruments. However, the wavetable synthesizer playing a short, looped sample of a natural instrument is easily distinguishable from the actual instrument. This results from the presence of specific modulatory structures in the sound of some instruments related to expression, which can be a strong clue regarding the identification of the instrument. The control of early electrophones, such as the theremin or Martenot waves, gave the performer expressive capabilities comparable to bowed instruments. Contemporary synthesizers are returning to similar solutions. The aim of this work is to study the impact of various types of modulation on the perceived naturalness of violin sound. Modulation through an automatic low frequency oscillator is compared to expressive modulation by a human using a controller. Two advanced controllers are studied to determine whether simultaneous modulation of more than one parameter brings benefits. A set of sound samples was prepared that included violin recordings and synthesized signals, where different waveforms were combined with various modulation sources and modulated parameters. The effect was assessed by a group of expert listeners. The results indicate that expressive, multi-parameter modulation with advanced controllers brings benefits for waveforms with realistic spectra, close to that of a violin. In less realistic waveforms this kind of modulation may be perceived as less natural than a simple one, obtained through an oscillator.