Blind Source Separation with Possibilistic Variables
Thaddeus T. Shannon · 2005
This paper proposes a method for blind source separation (BSS) of observed variables characterized by possibility distributions. Techniques for BSS with probabilistic variables have been developed over the last decade under the general heading of independent component analysis (ICA). This paper proposes an analogous approach for linear mixtures of real sources characterized by possibility distributions. The methodology seeks a linear transformation of the observed variables that minimizes the interaction between sources based on a Hartley-like function proposed by Yuan and Klir for measuring the nonspecificity of real variables.