Lp-norm of the LMS algorithm for bilinear forms with α-stable processes.

In this paper, a novel bilinear algorithm is proposed for multiple-input–single-output (MISO) system identification, which is based on a modified L p -norm cost function with fractional lower order statistics. To model the MISO system, we employ the bilinear form (BF) which is defined with respect to the impulse responses of a spatiotemporal model. Considering the non-Gaussian behavior in practical MISO system with BF, α -stable distribution, whose density function decays in the tails less rapidly than the Gaussian density function, is used to model the interference noise. As an added contribution, we extend the least mean pth power (LMP) and normalized LMP (NLMP) algorithms to BF, resulting in the LMP-BF and NLMP-BF algorithms. Simulation results demonstrate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]