Optimal binaural processing based on point process models of preprocessed cues
Johnson, Don H.
Master of Arts
In localizing high-frequency tonal signals in the horizontal plane, the lateral superior olive (LSO) encodes the interaural level differences (ILD), the crucial cue for azimuthal angle extraction at some nuclei more central than LSO in the binaural pathway. The current work focuses on optimal processing at higher center(s), with inputs from LSO and anteroventral cochlear nucleus (AVCN). Besides Poisson processes, non-Poisson point processes are developed to model these cell responses more accurately. The LSO projection (if available), is shown to be the only input used in maximum likelihood angle estimation, and it is processed similarly under different models of LSO responses. However, localization performance depends strongly on the model's regularity. A more regular process, consistent with LSO chopping responses, improves localization accuracy and precision. Temporal integration in lower nuclei and neuron ensemble processing are both helpful to binaural localization at this stage.
Electronics; Electrical engineering; Neurosciences; Statistics