Research On Robust Adaptive Networks

The robustness against impulsive interference is a significant research direction in the field of signal processing.The sign error least mean square?SE-LMS?algorithm provides an effective solution for addressing impulsive noise problem.However,when the input signal to the unknown system is colored,or the unknown system is sparse,the SE-LMS algorithm yields slow convergence rate and large steady-state misalignment.In Recent years,adaptive networks developed from adaptive filters have found wide applications in many fields,such as target location and tracking,movement modeling of animal groups,gene regulatory networks,spectrum sensing,sensor networks,and so on.One of the important factors influencing adaptive network performance is adaptive algorithms.This thesis firstly applies the idea of variable step-size into the diffusion affine projection sine algorithm?DAPSA?which has fast convergence rate and is robustness against impulsive noise,and proposes a variable step-size diffusion affine projection sign algorithm?VSS-DAPSA?,which can obtains both fast convergence rate and small steady-state misalignment.Then this thesis applies the methods of affine projection and sign operation into multitask networks,and presents a robust multitask affine projection sign algorithm?MD-APSA?.This algorithm is suited to multitask networks and has fast convergence rate and lower steady-state misalignment.Finally,this thesis applies the l₀–norm of the adaptive filter parameter vector into the cost function of the MD-APSA,and presents a sparse robust multitask affine projection sign algorithm?SMD-APSA?.The SMD-APSA can enhance the convergence rate when the estimated system is sparse.