LMS Coefficient Filtering for Time-Varying Chirped Signals

Abstract

This paper presents coefficient filtering techniques in the least mean squares (LMS) algorithm to improve adaptive predictor tracking performance for time-varying chirped signals. The example application used in this paper is an electronic support measure (ESM) receiver for detecting radar chirped pulses. The leakage LMS, momentum LMS, and the proposed future-state coefficient (FC-LMS) filtering algorithms have been studied. The leakage LMS algorithm has the ability to remove the memory effect of the initial converged time-varying frequency of the chirped signal, thus improving the radar pulse detection performance. The momentum LMS is able to search for the time-varying optimum weight solution more efficiently, and the FC-LMS uses a parallel technique to retain the LMS throughput while being able to show a better tracking performance for chirped signals compared with the standard LMS algorithm.