A computational model for amplitude modulation extraction and analysis of simultaneous amplitude modulated signals

Abstract
The existence of amplitude-modulation (AM) frequency-specific channels has previously been suggested on the basis of psychoacoustical experiments using an adaptation or masking paradigm. Electrophysiological studies have also shown that cochlear nucleus chop-s neurons are tuned to specific AM frequencies. We have developed a simple but physiologically plausible model of chop-S neuron properties based on a digital integration of Hodgkin and Huxley equations. Although it does not include an exhaustive peripheral processing, our model produces similar results to physiological data. We recover a bandpass AM transfer function at high input levels and non-linear filtering. This model enables us to simulate physiological masking situations in the modulation domain. We examine the strength of the model synchronous response to white noises modulated at two different frequencies and added with various intensity ratios. Simulations show that the chop-S model enhances AM separation of a well-tuned AM frequency added to another, relative to the auditory-nerve fibers inputs. Masking patterns computed by varying the masker-modulation frequencies show the recruitment of the neuron by these stimuli.