Abstract:In the past decade, the augmented complex-valued least mean square adaptive filtering algorithm based on the minimum mean square error criterion and widely-linear model has attracted much attentions due to its simplicity and applicability to non-circular signals. However, the system noise with impulse characteristics will dramatically affect the convergence of the augmented complex-valued least mean square algorithm. Recently, a robust widely-linear complex-valued least mean M-estimate algorithm has been proposed, which significantly improves the robustness of the augmented complex-valued least mean square algorithm to impulsive noise. But it should be noted that the widely-linear complex-valued least mean M-estimate algorithm with fixed step size faces the trade-off between convergence speed and steady-state accuracy. To overcome this drawback, a variable step size widely-linear complex-valued least mean M-estimate algorithm is presented in this paper, in which the optimal step size at each iteration is obtained by maximizing the difference between the mean square deviations of the WL-CLMM algorithm at adjacent moments. In order to reveal the statistical convergence behavior of the proposed variable-step-size widely-linear complex-valued least mean M-estimate algorithm to better guide the parameter selection in practical applications. Theoretical transient and steady-state mean-square deviation convergence behaviors of the variable-step-size widely-linear complex-valued least mean M-estimate algorithm are analyzed and the corresponding mean-square deviation expressions are also derived. Computer simulations on system identification and stereophonic acoustic echo cancellation in impulsive noise environments confirm the validity of the analysis results and the performance improvement from the variable step size strategy.

Collaborative Adaptive Volterra Filters for Nonlinear System Identification in Α-Stable Noise Environments

Identification of Nonlinear System with Irregularly and Slowly Sampled Output Using Lpv Model

A recursive adaptive lattice filters for impulsive signals

An Adaptive Linear Element Based Method for Identification of Linear Parameter Varying Models

Adaptive Kernel Leaning Networks with Application to Nonlinear System Identification

A robust adaptive decomposable Volterra filter based on the hyperbolic tangent Leclerc function and its performance analysis

Adaptive Hammerstein Filtering Via Recursive Non-Convex Projection

Parallel Nlms Filters With Stochastic Active Taps And Step-Sizes For Sparse System Identification

A New Adaptive Method for Nonlinear System Modeling

Affine Combination of the Filtered-x LMS/F Algorithms for Active Control

Kernel Least-Mean Mixed-Norm Algorithm

Reliable Filtering of Nonlinear Markovian Jump Systems: the Continuous-Time Case

Novel Improvement for Nonlinear Compatibility of Least Mean Square Adaptive Algorithm

Robust Widely Nonlinear Quaternion Volterra Adaptive Filter Based On Recursive General Barron Algorithm and Its Performance Analysis

Robust Variable Step Size Widely Linear Complex-Valued Least Mean M-estimate Adaptive Algorithm: Derivation and Performance Analysis

Optimal Volterra-based nonlinear system identification using arithmetic optimization algorithm assisted with Kalman filter

Two are Better Than One: Adaptive Sparse System Identification using Affine Combination of Two Sparse Adaptive Filters

Nonlinear System Identification Using Kautz Basis Expansion-Based Volterra–PARAFAC Model

Iterative Learning Identification with Bias Compensation for Stochastic Linear Time-Varying Systems.

L0-norm constraint normalized subband adaptive filtering algorithm: Performance development and AEC application

Regularized Least-Mean-Square Algorithms