A higher-order statistics-based adaptive algorithm for line enhancement

Abstract

A novel higher order statistics (HOS) based adaptive filtering algorithm for line enhancement is suggested. The enhancement process is achieved by filtering the noisy signal through an adaptive FIR filter. The steady state of the impulse response of this filter is proportional to a selected one-dimensional (1-D) slice of the fourth-order mixed cumulant of the input signal. It is shown that this slice is comprised of noiseless sinusoids if the input signal is comprised of sinusoids embedded in Gaussian noise. Therefore, the algorithm is considered to be a suitable one in processing sinusoids embedded in highly colored Gaussian noise. Simulation results verify the performance of the proposed algorithm.