Auditory analysis of spectro-temporal information in acoustic signals

Abstract

Acoustic signals are often characterized by pitch, timbre, loudness, forms of modulation, and onset/offset instants. These descriptors of sound quality have a close relationship to the instantaneous spectral properties of the sound waves. The auditory system has developed elegant mechanisms to extract and represent this spectro-temporal information through noise-robust perceptual features. After decades of research, the anatomical organization and functional significance of many of the underlying mechanisms are now becoming understood. Consequently, mathematical descriptions of the various early and central auditory stages have become feasible and functionally relevant, and can be fruitfully used to derive principles underlying time-frequency analysis in the auditory system. Here, the authors review examples of such approaches from their recent efforts to model the early and central stages of the mammalian auditory systems. It is within the hope of this study to better understand the nature of hearing, and to eventually apply these functional principles to designing engineering applications, such as audio coding algorithms that are closely optimized under perceptual criteria, and automatic speech recognition systems that have performance comparable to humans.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>