Basics of Auditory Processing

This talk reviews the signal processing strategies that have been discovered at the early and the cortical stages of the auditory nervous system. This system takes in as input a one-dimensional sound waveform, and then transforms it in a sequence of levels into various representations in order to extract a diverse array of sound attributes such as timbre, pitch, location, and loudness. It has been discovered that the first operation at the cochlea resembles closely a wavelet spectral transformation with special characteristics that are very useful in many audio and speech applications. At the cortical level, the cochlear spectral analysis is further elaborated by a two-dimensional spectro-temporal wavelet transformation which produces extremely rich cortical representations. These are then exploited by the brain to attend to sounds, to segregate complex mixtures, to provide for robust listening in noisy environments, and also to perform numerous cognitive functions with speech, music, and environmental sounds such as creating categories, making decisions, and extracting meaning.