Suppression and (2f1-f2)-difference tones in a nonlinear cochlear preprocessing model with active feedback.

Abstract

The two nonlinear effects of two-tone suppression and of (2f1-f2)-difference tone creation are measured in a hardware model which consists of 90 sections containing nonlinear feedback loops. The basic data are the level and phase distributions along the 90 sections produced by single tones in the linear passive system which are almost identical to those produced in the nonlinear active system at high levels. Enhancement is created at medium and low input levels resulting in more strongly peaked level-place patterns. Two-tone suppression is, therefore, described as a "de-enhancement" which is produced by the gain reduction in the saturating nonlinearity of the feedback loop in consequence of increasing input levels (that of the feedback loop in consequence of increasing input levels (that of the suppressor as well!). Characteristics of suppression are given in normalized form. The creation of (2f1-f2)-difference tones is based on the same nonlinear effects. In each section, difference-tone wavelets are created which travel--changing level and phase thereby--to their characteristic place, where they add up to a vector sum corresponding to the audible difference tone. In case of cancellation, the vector sum has to be compensated by an additional tone of the same frequency and amount but opposite phase. Based on this strategy of (2f1-f2)-difference tone development, the relevant relations are measured on the model and averaged either in normalized graphs or in equations in order to offer the possibility to simulate the hardware model on the computer. Psychoacoustically measured cancellation data are compared with data measured using the model. The two data sets agree not only in general but also in many details indicating that the model describes cochlear nonlinear preprocessing to a useful approximation.