LMS adaption with hard-limited gradient

摘要

The authors present a novel normalization technique for the least-mean-square (LMS) algorithm that has practical application to adaptive array configurations. Different normalization techniques are reviewed and empirical results presented. Simulation results indicate that the algorithm's steady-state performance is nearly the same as that of the LMS, HLLMS (hard-limited LMS), and NLMS (normalized LMS), but that its transient response exhibits the same independence to total interference power as does the NLMS algorithm. The GNLMS (gradient NLMS) algorithm can be used in single-port, as well as multiport, array configurations. The GNLMS is the only way to make response time independent of input power using a single-port technique. Improvement in stability and some improvement in steady-state performance were also realized when the GNLMS algorithm was used in simulations of a single-port configuration.