> The underlying distribution from which a non-stationary signal arises > changes in time. Any distribution you estimate with a histogram reflects > only the 'already' observed signal, but may not reflect the distribution of > values in the future.
I'm trying to do a bayesian source separation, for which I need a prior on the source signals. I'd like to estimate that prior from other signals that are very much related to my source signals. Maybe, for this application, it might be sufficient to start with a histogram of the observed signal? Geert > ----- Original Message ----- > From: "Geert Verdoolaege" <[EMAIL PROTECTED]> > To: <[EMAIL PROTECTED]> > Sent: Saturday, March 13, 2004 5:28 AM > Subject: [edstat] distribution of non-stationary signal > > > > How can I estimate the probability distribution of a non-stationary > signal? > > > > Thanks, > > Geert > > > > > > . > > . > > ================================================================= > > Instructions for joining and leaving this list, remarks about the > > problem of INAPPROPRIATE MESSAGES, and archives are available at: > > . http://jse.stat.ncsu.edu/ . > > ================================================================= > > > > . > . > ================================================================= > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > ================================================================= . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
