Abstract
The human auditory system is capable of perceiving the direction of a sound source despite reflections off of walls, ceiling and other interfering objects. This phenomenon, the precedence effect, has been previously modeled only for synthetic click stimuli. This paper presents a biologically plausible computer simulation for speech, that combines a cross-correlation model with an onset enhancement scheme. As part of the complete model, a novel onset enhancement method is described and implemented, based on adaptive prediction. The adaptive method is promising in its ability to enhance onsets and reduce the steady state portion of the speech signal. The model can serve as a tool to understand the precedence effect for complex signals and should be advantageous to engineering applications such as preprocessors for speech recognition systems.