A speech signal processing system based on a peripheral auditory model

The theory of acoustic invariance states that phonetic features are decoded by the auditory system from complex invariant properties of the speech waveform. We describe a method of transforming DFT frames of a speech signal into a representation that models the response of the peripheral auditory system with special attention to the auditory system´s response to signals exhibiting rapid frequency and amplitude changes (consonants). The method involves three transformations, one in the frequency dimension, one in the amplitude dimension, and one in the temporal dimension (time course of auditory response). We describe each transformation, and show that invariant auditory response may be elicted by apparently disparate acoustic patterns.