An Auditory Scene Analysis Approach to Monaural Speech Segregation

Abstract

A human listener has the remarkable ability to segregate an acoustic mixture and attend to a target sound. This perceptual process is called auditory scene analysis (ASA). Moreover, the listener can accomplish much of auditory scene analysis with only one ear. Research in ASA has inspired many studies in computational auditory scene analysis (CASA) for sound segregation. In this chapter we introduce a CASA approach to monaural speech segregation. After a brief overview of CASA, we present in detail a CASA system that segregates both voiced and unvoiced speech. Our description covers the major stages of CASA, including feature extraction, auditory segmentation, and grouping.