Hyperacuity in time: a CNN model of a time-coding pathway of sound localization

This paper discusses a new multilayer one-dimensional (1-D) cellular neural network model of the time-coding pathway of sound localization. The key feature of the model is lateral inhibition, which is supposed to play a crucial role in sound localization. The possible role of this inhibition is examined on the basis of our model and several conclusions are drawn concerning the expected nature of inhibition. It is also shown that by use of inhibition, a group of neurons may be much more sensitive to interaural time difference than one individual neuron. Thus, our model of the first stage of the sound localization system solves a hyperacuity in time problem. The second part of the paper introduces a CNN model of that part of the sound localization system which is characterized by a massive convergence of different frequency channels to resolve the so-called phase ambiguity problem. We show that with inhibition good results can be achieved here too. Quantitative studies show the robustness of the model