FIG. 2.

A schematic illustration of the STMI based on the processing of AN neurograms by a bank of cortical spectro-temporal modulation filters producing clean “template,” T, and “noisy”, N, auditory cortex outputs. The NSIM is also illustrated and is based on the clean “reference,” R, and “degraded,” D, neurograms. In this study, unprocessed speech signals are applied to the models to produce the “reference” neurograms and “template” cortical responses, and chimaeric speech signals are applied to the models to obtain the “degraded” neurograms and “noisy” cortical responses. The optional LIN processing extracts additional information from the neurograms by accounting for phase offsets in the AN fiber responses.