An Adaptive Filtering Algorithm of Higher-Order Cumulant-Based Signed Coherent Integration

Traditional LMS (Least Mean Square) based ALE (Adaptive Line Enhancement) algorithm has three disadvantages: ability to hand Gaussian colored noise is bad, computational complexity is high, and noise variance of the output is great. For greatly reducing these three disadvantages, firstly, we used the integrated function of the coherent components of the input and the low computational load of the signed algorithm to modify the traditional LMS algorithm. This modified LMS algorithm is regarded as the signed coherent integration(SCI) algorithm. Secondly, we developed HOCSCI (higher-order cumulant signed coherent integration) algorithm for adaptive spectrum enhancement. The performance of the new algorithm is better than that of higher-order cumulant iteration (HOCI) algorithm. Compared with the HOCI and SCI algorithm, the new algorithm has the following features: (1) more signed coherent integrated terms are introduced into the cumulant updating equation. Thus it is easier to guarantee the adaptive integration action, and then very weak non-linear frequency modulation signals can be enhanced; (2) the computational load of the HOCSCI algorithm is smaller than that of the HOCI algorithm; (3) the stability of the output signals of the HOCSCI algorithm is better than that of the HOCI or SCI algorithm. Simulation results validate these conclusions.