Subject: Re: Blind Source Separation by Sparse Decomposition From: "Mikael.Fernstrom" <mikael.fernstrom(at)UL.IE> Date: Wed, 8 Sep 1999 12:03:17 +0100Rich & Michael, When clipping a signal as suggested you get some of the spectral componen= ts "re-coded" as pulse-width modulation. Although the resulting signal is spectrally substantially different, the variation in spectral "flux" by t= ime (especially with speech) is the same, hence we are probably talking about= a Gibsonioan invariant in this case. /Mikael -------------------------------------------------------------------- Mikael Fernstr=F6m, M.Sc. Interaction Design Centre & UL Interactive Media Group, Foundation Building, University of Limerick, Ireland Phone: +353-61-202606, Fax: +353-61-202572 Email: mikael.fernstrom(at)ul.ie Web: www.idc.ul.ie