THE RELATIVE STRENGTH OF THE TONE AND NOISE COMPONENTS IN ITERATED RIPPLED NOISE

Rippled noise is constructed by delaying a random noise and adding it back to the original. Iterated rippled noise (IRN) is constructed by r epeating the delay-and-add process. IRN produces a two-component perce ption, i.e., a buzzy tone with a pitch equal to the reciprocal of the delay and a background noise that sounds like the original random nois e. The perceived tone/noise ratio increases with the number of iterati ons. The effective tone/noise ratio in IRN sounds with 1-16 iterations was measured in a discrimination matching experiment: each IRN was pa ired with a range of standard sounds, having varying proportions of a broadband noise and a complex tone, to find the point where their perc eived tone/noise ratios are the same. The experiment shows that the to ne/noise ratio of the matching standard increases 3.8 dB per doubling of the number of iterations in the IRN stimulus. Spectral models of au ditory perception explain the pitch of IRN in terms of peaks in the re ion of the first five to eight harmonics of the reciprocal of the del ay. However, the matching data are unaffected when the sound is high-p ass filtered at the twelfth harmonic of the delay-above the region of resolved harmonics. We show that a wide range of time-domain auditory models can explain the discrimination matching data by applying autoco rrelation, either to the IRN waveform, or to the neural activity patte rns produced by the cochlea in response to IRN waves. (C) 1996 Acousti cal Society of America.