ABSTRACT

Given that the number of the signals M has been estimated beforehand and the original stochastic matrix U(0) is a N ( )N M− dimension matrix, the U(t) will approximate the noise eigen subspace of the sampled signals. Here the μ is called step factor. The iteration expression (6) is stable when the μ meets 0 < μ << 1/ maxλ , and the λmax indicates the maximum eigen value of the covariance matrix R(t).