Subject: Pitch-tracking From: David Huron - Conrad Grebel <dhuron(at)WATSERV1.UWATERLOO.CA> Date: Fri, 25 Feb 1994 23:21:25 -0500
Regarding pitch-tracking. Andy Moorer's very fine dissertation (1975) notwithstanding, the best work I have seen in this area is an Australian dissertation by Charles Watson. The full reference is: Charles Richard Watson, The Computer Analysis of Polyphonic Music. PhD dissertation, Basser Department of Computer Science, University of Sydney, 1985; 214pp. Watson's work has languished in obscurity (I have never seen a reference to it) -- yet it is superb. There are good descriptions of various technical approaches, and excellent evaluations of alternative pitch-extraction algorithms. The dissertation uses four movements from a trio for two oboes and bassoon as a benchmark test. Also there are good discussions of the pertinent perceptual literatures in event perception. Watson's work attends to both the problem of pitch extraction in polyphonic contexts, as well as the problem of duration extraction. His duration extractor algorithm is especially ingenious. Watson ran computationally obese programs on 1985-vintage mainframes. Modern computers would make his methods more tractable.