The effects of spectral fusion on the perception of pitch for complex tones

Abstract

Pitch can be shown to be a perceptual attribute that is built up from the internal representation of an actual or virtual “source” rather than being derived directly from the waveform. Michael Mc Nabb and John Chowning at CCRMA, Stanford University, have developed a technique whereby several sine wave components can be made to fuse into a unitary percept by introducing parallel vibrato (combined random and periodic frequency modulation at subaudio frequencies) to all spectral components. If different subsets of components in a spectral complex are subjected to different vibrato patterns, they separately fuse into virtual sources, each with their respective pitch. Spectral fusion depends on the harmonicity and recognizability of the spectral subset, e.g., whether it resembles a voice, trumpet, etc. And the separation of subsets depends, in part, on the correlation of their vibrate functions. With inharmonic spectra, pitch changes can be made to occur by causing different parsings of the spectrum into sources. This strongly suggests that pitch is a property assigned to a source after a certain amount of perceptual and cognitive processing of the time-arying spectrum has taken place. Further, the fact that pitch depends on context, i.e., depends on the determination of the source on the basis of spectral congruity, necessitates a re-evaluation of the methodology of pitch research.