A biologically inspired binaural approach to monaural modeling

The auditory system is often discussed as having monaural and binaural neurological pathways; similarly models are classified as either monaural or binaural. Psychophysical evidence of contra-aural interference (when performance with one ear is better than performance with two ears) suggests that the information used on monaural tasks (e.g., N0S0 and NmSm detection) may be carried by a binaural pathway. Binaural models often require monaural channels to predict the results of monaural tasks, but these monaural channels prevent the models from predicting contra-aural interference. This modeling work investigates the monaural information carried by a processor which is inherently binaural. The processor design makes the inclusion of monaural channels unnecessary and contra-aural interference is predicted under certain conditions. The performance of the model matches results from a variety of traditional psychophysical tasks (including discrimination of differences in overall intensity; discrimination of differences in interaural level, time and coherence; as well as detection under monaural and binaural masking conditions). Results suggest that binaural neurons contain sufficient information to explain performance on both binaural and monaural tasks. [Work supported by NIH grants R01 DC 00100 and 1 F31 DC006769-01.]