signal processing - Robust least squares in Octave -
i programming classical prony's method (or prony's analysis) using octave. script works pretty fine on test signal such as:
t = 1:600; sig = exp(-0.01*t).*cos(0.1*t);
but totally collapses when small amount of noise added:
sig = sig + 0.001*rand(size(sig));
the issue computing overdetermined set of equations using pinv
when linear prediction bulit (see link above). need more robust , less sensitive outliers. sadly, no function in octave core helped me far. have tried ols
, lscov
don't trick.
any hints, please?
note: aware classic prony problematic because of issues , there modified algorithms (such this one). feel haven't done maximum classic method work. using better solver should persist kind of noise.
Comments
Post a Comment