![]() |
Workshop on the Auditory Basis of Speech PerceptionKeele University, UK |
![]() |
I address the question of source separation using low-level knowledge about signals. Assuming that we are able to well estimate characteristics of the signals, like fundamental frequency (f0) and interaural time differency (ITD), I show how to use these measures in order to directly recover the signal produced by a given source. This ideal f0-guided segregation process well works with quasi-stationary signals, with no or well-described f0 variations. It enables to recover periodic signals embedded in noise and to separate double-vowels. I discuss widely the place of this new algorithm in the context of auditory scene analysis (ASA).
Bibliographic reference. Berthommier, Frédéric (1996): "Direct separation of sounds based on knowledge of F0 and ITD", In ABSP-1996, 170-173.