On the performance of the MLE of adaptive array weights: a comparison

Abstract

The maximum-likelihood (ML) adaptive weight vector is derived, and its performance is compared with those of the sample matrix inversion (SMI) method and the least-mean-square (LMS) algorithm. The ML adaptive weight vector and the SMI method achieve identical convergence rates for the average SNR. With the desired signal absent, they are superior to the LMS algorithm. However, when the desired signal is present and the optimum SNR which depends on the received SNR is large, they lose their superiority. For the average MSE performance, the convergence of the ML adaptive weights is the fastest when the optimum SNR is high enough. With a small optimum SNR, the SMI method performs better than the other algorithms.<>