Timbre and pitch strength in iterated rippled noise

Abstract

Iterated rippled noise (IRN) produces a perception with both tonal and noisy components, even when high-pass filtered to remove all resolved harmonics. IRN stimuli are generated by cascading ‘‘delay and add’’ networks, where the delay, gain after delay, and number of iterations are the variables that determine the pitch and the perceptual tone/noise ratio. The strength of the tonal component is directly related to the degree of regularity in the fine structure; the strength of the noisy component is inversely related to regularity. Autocorrelation, correlograms, and summary correlograms provide quantitative measures of regularity that explain the pitch and the perceptual tone/noise ratio of IRN stimuli. Experiments with IRN stimuli involving pitch matching, pitch strength discrimination, and masking show that temporal correlation measures provide better explanations of the data than do spectral measures. The perceptual difference between the tonal and noisy components is a timbre distinction; it is also referred to as relative pitch strength. IRN is also an example of streaming in the absence of spectral cues—streaming based on temporal regularity. The main properties of IRN will be illustrated with correlogram videos. [Work supported by NIDCD and DRA Farnborough.]