Auditory model inversion and its application

Auditory model has been applied to several aspects of speech signal processing field, and appears to be effective in performance. This paper presents the inverse transform of each stage of one widely used auditory model. First of all it is necessary to invert correlogram and reconstruct phase information by repetitious iterations in order to get neural firing rate. The next step is to obtain the negative parts of the signal via the reverse process of the HWR (Half Wave Rectification) stage. Finally the functions of inner hair cell/synapse model and Gammatone filters have to be inverted. Thus the whole auditory model inversion has been got. An application of noisy speech enhancement based on auditory model inversion is mentioned in this paper. Many experiments show that this method is effective for noisy speech, especially for low SNR. The method of auditory model inversion given in this paper could be used widely in speech enhancement field.