Application of loudness models to sound processing for cochlear implants

A new paradigm for processing sound signals for multiple-electrode cochlear implants is introduced, and results are presented from an initial psychophysical evaluation of its effect on the perceived loudness of complex sounds. A real-time processing scheme based on this paradigm, called SpeL, has been developed primarily to improve control of loudness for implant users. SpeL differs from previous schemes in several ways. Most importantly, it incorporates a published numerical model which predicts the loudness perceived by implant users for complex patterns of pulsatile electric stimulation as a function of the pulses’ physical parameters. This model is controlled by the output of a corresponding model that estimates the loudness perceived by normally hearing listeners for complex sounds. The latter model produces an estimate of the specific loudness arising from an acoustic signal. In SpeL, the specific loudness function, which describes the contribution to total loudness of each of a number of frequency bands (or cochlear positions), is converted to a pattern of electric stimulation on an appropriate set of electrodes. By application of the loudness model for electric stimulation, this pattern is designed to produce a specific loudness function for the implant user which approximates that produced by the normal-hearing model for the same input signal. The results of loudness magnitude estimation experiments with five users of the SpeL scheme confirmed that the psychophysical functions relating overall loudness perceived to input sound level for five complex acoustic signals were, on average, very similar to those for normal hearing.