Complex independent component analysis by nonlinear generalized Hebbian learning with Rayleigh nonlinearity

This paper presents a non-linear extension of the Sanger´s (1989) generalized Hebbian algorithm to the processing of complex-valued data. A possible choice of the involved nonlinearity is discussed recalling the Sudjianto-Hassoun (1994) interpretation of the nonlinear Hebbian learning. Extension of this interpretation to the complex case leads to a nonlinearity called the Rayleigh function, which allows for separating mixed independent complex-valued source signals