Research On Sparse Adaptive Filtering Algorithms

Sparse systems exist widely in areas of communications and information processing.In recent years,the research on the topic of sparse system estimation has attracted much attention.At present,some sparse adaptive filtering algorithms can address the problem of sparse system estimation well under general conditions.However,in some special cases,such as the scenarios where exist impulsive interference or where the adaptive filter is operating with complex signals,the performances of the algorithms are not ideal enough or the algorithms cannot be used.In order to address the problem above,this thesis firstly applies the idea of sign operation to adaptive estimation of block sparse systems and proposes a block sparse sign algorithm(BS-SA).This algorithm can effectively restrain the influence of impulsive noise on the performance of the adaptive filter.Then this thesis analyzes the complex zero-attracting least mean square(ZA-CLMS)algorithm for circular Gaussian input signal.The result of performance analysis can predict the estimation accuracy of the algorithm when the adaptive filter arrives at steady state.Finally,this thesis extends the idea of sign operation to complex domain.By defining a weighted cost function of the estimation error and using the proportionate adaption strategy in the method of gradient descent,a robust complex proportionate normalized sign algorithm(CIPNSA)is derived.This algorithm can both obtain good robustness against impulsive interference and can speed up the convergence rate of the adaptive filter estimating complex sparse system.