MODELLING AUDITORY NEURONES WITH HODGKIN AND HUXLEY EQUATIONS — Application to Amplitude Modulation Processing

Physiological recordings have shown that stellate cells in the ventral cochlear nucleus (VCN) enhance amplitude modulation (AM) information, relative to their auditory nerve (AN) inputs. Modelling spike-discharge properties of VCN neurones implies to take into account the stochastic aspect of AN fibres firings and the fine temporal aspects of the neural integration occurring at the level of the VCN. These two aspects can be simulated using respectively independent Poisson point processes and a digital integration of Hodgkin and Huxley equations. We determine the response transfer functions to AM, using sinusoidally AM white noises. The resulting modulation transfer functions (MTF) based on spike count is all-pass while the computed MTF based on modulation gain is low-pass at low intensity and bandpass at high intensity. These results are closely compatible with real data and confirm that AM encoding is performed by good response synchrony at preferential AM rates. A detailed study of the nonlinear MTF also predicts two secondary preferential frequencies harmonically related to the primary one. We finally examine the model MTF in the presence of a continuous wideband background noise. AM coding performed by the stellate cell model is robust as simulations show enhanced selectivity in the presence of significant levels of wideband background noise.