Using a combined localization/detection model to simulate human localization performance near the masked detection threshold level

Abstract

Recently, the perceptual lateralization of a partly masked target has been successfully simulated using the interaural cross-correlation difference (ICCD) model [Braasch, J. Acoust. Soc. Am. 108, 2597 (2000)]. However, in the accompanying listening tests, the target level was often found to be below the masked detection threshold level. To improve the model performance, a detection threshold model has been implemented in the localization model so that the localization process is triggered only when the target level is above a threshold. Otherwise, the position of the sound source is determined according to a behavioral pattern. The detection stage is based on an on- and offset detection algorithm that analyzes the derivative in time. For the simulation of the binaural conditions, the equalization-cancellation algorithm of Durlach [J. Acoust. Soc. Am. 35, 1206–1218 (1963)] has been included in the model. It has been shown that for target levels above the masked detection threshold, the on-/offset detection algorithm is accurate enough to trigger the subtraction process of the ICCD algorithm. Furthermore, the improved model has resulted in better simulations of human localization patterns in the presence of a distracting sound.