A precedence effect model to simulate localization dominance using an adaptive, stimulus parameter-based inhibition process.

A number of precedence-effect models have been developed to simulate the robust localization performance of humans in reverberant conditions. Although they are able to reduce reverberant information for many conditions, they tend to fail for ongoing stimuli with truncated on/offsets, a condition human listeners master when localizing a sound source in the presence of a reflection, according to a study by Dizon and Colburn [J. Acoust. Soc. Am. 119, 2947-2964 (2006)]. This paper presents a solution for this condition by using an autocorrelation mechanism to estimate the delay and amplitude ratio between the leading and lagging signals. An inverse filter is then used to eliminate the lag signal, before it is localized with a standard localization algorithm. The current algorithm can operate on top of a basic model of the auditory periphery (gammatone filter bank, half-wave rectification) to simulate psychoacoustic data by Braasch et al. [Acoust. Sci. Tech. 24, 293-303 (2003)] and Dizon and Colburn. The model performs robustly with these on/offset truncated and interaural level difference based stimuli and is able to demonstrate the Haas effect.