Cascaded adaptive canceller using loaded SMI

Abstract

A fast-converging, highly parallel/pipeline cascaded canceler which uses the 2-input loaded sample matrix inversion (SMI) algorithm as the fundamental building block is developed which has convergence performance almost identical to one of the standards of a fast-converging adaptive canceler, the fast maximum likelihood (FML) canceler. Furthermore, the new algorithm, denoted as the cascaded loaded SMI (CLSMI), does not require the numerically intensive singular value decomposition (SVD) of the input data matrix as does the FML algorithm. For both the FML and CLSMI developments it is assumed that the unknown interference covariance matrix has the structure of an identity matrix plus an unknown positive semi-definite Hermitian (PSDH) matrix. The identity matrix component is associated with the known covariance matrix of the system noise and the unknown PSDH matrix is associated with the external noise environment. For narrowband (NB) jamming scenarios with J jammers it was shown via simulation that the CLSMI and FML converge on the average -3 dB below the optimum in about U independent sample vectors per sensor input. Both the CLSMI and FML converge much faster than the standard canceler technique, the SMI algorithm.