A new procedure for automatic fitting of the basilar-membrane input-output function to individual behavioral data.

The basilar membrane input-output function (BM I/O) in a healthy cochlea is highly nonlinear. One of the consequences of sensorineural hearing loss (SNHL) is a partial or full loss of this nonlinearity. Behavioral estimates of the individual BM I/O can be useful for modeling the impaired auditory system and, potentially, for clinical diagnostics. Computational algorithms are available that mimic the functioning of the nonlinear cochlear processing. One such algorithm is the dual resonance non-linear (DRNL) filterbank [6]. Its parameters can be modified to account for individual hearing loss, e.g., based on behavioral, temporal masking curves (TMC) data. This approach was used within the framework of the computational auditory signal-processing and perception (CASP) model to account for various aspects of SNHL [4].
However, due to the computational complexity, on-line fitting of the DRNL parameters is difficult.
Until recently, the parameters were manually adjusted and the fitting process was indirect. A new approach is described here, based on a search through a lookup table of pre-computed filterbank
input-output functions. The aim of this approach is to provide a fast, stable, and more objective fitting procedure.