Lyon's Auditory Model Inversion: a Tool for Sound Separation and Speech Enhancement

A new implementation of Lyon's Auditory Model and an optimised inversion procedure will be presented. Both the passive and active Lyon's cochlea models were studied as new signal processing analysis schemes, while only the first one was considered regarding the inversion procedure. Following the work of M. Slaney, sound resynthesis was obtained inverting the correlogram representation by a new optimised algorithm. The utility of auditory model inversion will be emphasised focusing on the problem of speech enhancement and sound separation.