Tracking performance of leakage LMS for chirped signals

Abstract

The initial frequency pass-band of the LMS filter remains whilst tracking a non-stationary chirped signal. This past memory effect causes unwanted white noise to leak through the initial residual pass-band of the adaptive filter. A leakage term is applied to the LMS algorithm to remove the memory effect of the tracking filter which leads to a reduction in the noise power at the output of the adaptive filter. This reduced noise power is reflected in an improved SNR (signal-to-noise ratio) of a low SNR chirped signal compared to the standard LMS algorithm.