Recursive blind LMS parameter identification for single-input multiple-output system

A blind least-mean-squares (BLMS) algorithm is proposed for the parameter identification of single-input multiple-output (SIMO) systems. Without requiring knowledge of a reference signal, it is proved that the presented parameter estimates almost sure converge to their real value under the assumption that the observed noises are mutually independent distributed additive white sequence with known variance. The noise variance of observed signal is estimated from the eigenvalues of a matrix related to observed signal sequence firstly. We back our theoretical findings with experiments showcasing the potential merits of the BLMS in practice.