Scholarlink

<p>In this paper, a mixed generalized correntropy adaptive filter with jointly learned parameters is proposed for robust adaptive filtering in impulsive non-Gaussian noise environments. The mixture combines different correntropy kernels with the generalized correntropy kernel, where the shape parameter <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\alpha \)</EquationSource> <EquationSource Format="MATHML"><math> <mi>α</mi> </math></EquationSource> </InlineEquation> controls the tail behavior of each kernel. By simultaneously learning the dispersion parameters and the mixture weight online through a steepest-descent rule, the proposed method provides enhanced performance. Furthermore, mean and mean-square convergence analyses for both mixed correntropy and mixed generalized correntropy are derived, and explicit step-size stability bounds are established. Simulation results confirm the superior performance of the proposed algorithm compared to the existing methods, including fixed-parameter correntropy filters and classical robust baselines.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Robust Mixed Generalized Correntropy Adaptive Filter with Learned Parameters: Mean and Mean-Square Convergence Analysis

  • 期刊论文

  • 发表日期：

    2026年4月13日

查看全文
  • Mohammad Salman,
  • Hadi Zayyani,
  • Hasan Abuhilal,
  • Mostafa Rashdan

摘要

In this paper, a mixed generalized correntropy adaptive filter with jointly learned parameters is proposed for robust adaptive filtering in impulsive non-Gaussian noise environments. The mixture combines different correntropy kernels with the generalized correntropy kernel, where the shape parameter \(\alpha \) α controls the tail behavior of each kernel. By simultaneously learning the dispersion parameters and the mixture weight online through a steepest-descent rule, the proposed method provides enhanced performance. Furthermore, mean and mean-square convergence analyses for both mixed correntropy and mixed generalized correntropy are derived, and explicit step-size stability bounds are established. Simulation results confirm the superior performance of the proposed algorithm compared to the existing methods, including fixed-parameter correntropy filters and classical robust baselines.