For both normal hearing subjects and cochlear implantpatients the most drastic step of sound coding for neuronalprocessing is when the analog signal is converted intodiscrete nerve-action potentials. As any information lostduring this process is no longer available for neuralprocessing, it is important to understand the underlyingprinciples of sound coding in the intact auditory systemand the limitations in the case of direct electricalstimulation of the auditory nerve.Here we focus on a model of spiral ganglion type I neuronswith Hodgkin-Huxley type ion channels, which are alsofound in cochlear nucleus neurons (KA, Kht, Klt). Dependingon the task, we model the neurons at different levels ofdetail. For acoustic stimulation, we model the postsynapticbouton (1.5 x 1.7 µm) from high-spontaneous rate fibersand a synaptic excitation model fitted to results fromGlowatzki and Fuchs (2002). To study the response toelectrical stimulation in detail, we use a multi-compartmentmodel and an approximation of the electrical field along theneuron. For automatic speech recognition and informationtheoretic calculations we use simplified singlecompartment versions.We analyze the quality of coding with the framework ofautomatic speech recognition and the methods ofinformation theory. Our results show that for acousticstimuli, the model provides realistic refractoryness andgenerates more realistic spike trains compared to anartificial spike generator. Not surprisingly, speechdiscrimination in electrical hearing is lower than in acoustichearing. This is probably due to the limited dynamic rangeof electric hearing and the wide current spread, whichlimits spectral resolution. On the other hand, the temporalprecision of information coding seems to be very highbecause at levels well above threshold, action potentialsare elicited quasi deterministic by the electrical stimuli. Weargue that CIS strategies a) waste as much as 50     «

For both normal hearing subjects and cochlear implantpatients the most drastic step of sound coding for neuronalprocessing is when the analog signal is converted intodiscrete nerve-action potentials. As any information lostduring this process is no longer available for neuralprocessing, it is important to understand the underlyingprinciples of sound coding in the intact auditory systemand the limitations in the case of direct electricalstimulation of the auditory nerve.Here we focus on a model o...     »