The precedence effect for noise bursts of different bandwidths. II. Comparison of model algorithms

Abstract

In this investigation, different model algorithms were tested for their ability to simulate the precedence effect for ongoing (non-impulsive) noise-bursts of different bandwidths (100 Hz, 400 Hz and 800 Hz). The psychoacoustical reference data—in which the perceived lateral position of a noise burst (200-ms duration, 20-ms cos 2 -ramps, 500-Hz center frequency) in the presence of one reflection (inter-stimulus interval: 0.0 ms-0.4 ms) was determined—were taken from a preceding paper on this investigation. It is shown that models which simulate the precedence effect by using the special characteristics of the auditory periphery or by focusing on the spectral dominance region fail when stimuli of longer duration than clicks are used, while a modified Lindemann model still shows satisfactory results. Furthermore, it was found that the trading ratio of ITDs and ILDs can be assumed to be constant for the stimuli tested. A discounting of ITD cues, as was found by Rakerd and Hartmann (1985), was not observed for the type of stimuli tested here. Instead, a discounting of the precedence effect occurred for some of the listeners when the bandwidth of the signals was very narrow. In the model simulation, it was not necessary to consider cross-frequency-band interaction like the second coincidence weighting of Stern et al. (1988), and it was sufficient to estimate the average of the outputs of the involved frequency bands.