A nonlinear filter bank model for the auditory periphery

The intrinsically nonlinear character of auditory processing has long been the bane of linear models for the auditory periphery. A previous computational model [Carney, J. Acoust. Soc. Am. 93, 401–417 (1993)] for single auditory nerve fibers contained a feedback mechanism to vary the bandwidth of a narrow-band gammatone filter as a function of sound pressure level. That model simulated the compressive nonlinearity, including aspects of temporal and average discharge rate characteristics of low-frequency auditory nerve fiber responses. The proposed model extends the previous work; its architecture is motivated by the most recent empirical evidence on outer hair cell (OHC) physiology. The nonlinear narrow-band gammatone filter is replaced with a broad second-order bandpass filter [G. von Békésy, J. Acoust. Soc. Am. 21, 245–254 (1949)] inside a feedback loop. The feedback path consists of a nonlinearity representing OHC transduction, followed by a lowpass filter for the cell membrane, and finally a nonlinearity representing OHC motility. A bank of such models allows representation of spatio-temporal discharge patterns across populations of auditory nerve fibers. [Work supported by The Whitaker Foundation and NIDCD.]