ABSTRACT
Pitch is a potentially critical component of an auditory scene analysis system. It is therefore important that the pitch extraction system is robust and flexible. An algorithm is outlined which is faithful to psychophysical phenomena. It has been successfully used to segregate simultaneous vowels by emphasizing frequency-selective channels on the basis of pitch signatures. Two recent challenges to the pitch extraction algorithm are addressed. These suggest that there are at least two different pitch segregation mechanisms; one is accurate, phase insensitive and uses low harmonics; the other is inaccurate, is phase sensitive and uses high harmonics. Explorations using the model show that it is able to account for these new data without amendment. The second challenge involves a suggestion that pitch may normally be used to exclude channels rather than select them is accepted as plausible.
