Geometrical structures of FIR manifold and multichannel blind deconvolution (Q1597619)

The geometrical structures of the manifold associated with finite impulse response (F.I.R.) filters are studied, and a natural gradient learning algorithm for blind deconvolution is developed. A Lie group structure is introduced, and the Riemannian metric is derived. A learning algorithm that is based on the natural gradient approach and its stability are analyzed. The problem deals with signal processing without knowing the source signals and the mixing system and using only the sensor signals to estimate the original source signals.