Spectral Resolvability of Iterated Rippled Noise

Abstract

A forward-masking experiment estimated the spectral ripple of iterated rippled noise (IRN) that is possibly resolved by the auditory system. Tonal signals were placed at spectral peaks and valleys of IRN maskers for a wide variety of IRN conditions that included different delays, number of iterations, and stimulus durations. The differences in the forward-masked thresholds of tones at spectral peaks and valleys were used to estimate spectral resolvability, and these results were compared to estimates obtained from the Auditory Image Model (AIM). IRN spectra have spectral peaks that are harmonics of the reciprocal of the delay used to generate IRN stimuli. As the number of iterations in the generation of IRN stimuli increases, so does the difference in the spectral peak-to-valley ratio. For high number of iterations, long delays, and long durations evidence for spectral resolvability existed up to the 5th harmonic. For all other conditions spectral resolvability appeared to disappear at harmonics lower than the 5th, or was not measurable at all. In contrast, peaks in the Summary Correlograms are clearly present for all of the IRN stimuli allowing them to predict the pitch of the IRN stimuli.