Scholarlink

<p>This work focuses on the design of adaptive filtering algorithm under censored observations and non-Gaussian impulsive interference. In practice, sensor saturation or unreliable communication usually leads to censored observations, in which case traditional adaptive filtering algorithms will produce significant estimation biases. Meanwhile, non ideal noise environments (non-Gaussian and impulsive noise) can induce severe performance degradation or even divergence of adaptive filtering algorithms based on error second-order moment optimization. To overcome these drawbacks, a censored regression maximum versoria criterion (CR-MVC) algorithm is proposed in this paper. Specifically, it describes the censored observation situation through a probit regression model and uses the maximum versoria criterion to suppress the adverse effects of non-Gaussian impulsive noise on the update of filter coefficients and background noise variance, ultimately achieving high steady-state estimation accuracy at a lower computational cost. Under common assumptions, the stability and steady-state mean square deviation (MSD) of the CR-MVC algorithm are analyzed, and the stability step size conditions and MSD theoretical model are also derived. Simulation experiments under censored observations and non-Gaussian impulsive noise environment are provided, confirming the accuracy of the theoretical MSD model and the superiority of the CR-MVC algorithm over other competing algorithms.</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 maximum versoria criterion adaptive filtering algorithm against non-Gaussian noises and censored observation

  • 期刊论文

  • 发表日期：

    2025年9月11日

查看全文
  • Shaohui Lv,
  • Wenjing Xu,
  • Jie Shi,
  • Pucha Song,
  • Haiquan Zhao

摘要

This work focuses on the design of adaptive filtering algorithm under censored observations and non-Gaussian impulsive interference. In practice, sensor saturation or unreliable communication usually leads to censored observations, in which case traditional adaptive filtering algorithms will produce significant estimation biases. Meanwhile, non ideal noise environments (non-Gaussian and impulsive noise) can induce severe performance degradation or even divergence of adaptive filtering algorithms based on error second-order moment optimization. To overcome these drawbacks, a censored regression maximum versoria criterion (CR-MVC) algorithm is proposed in this paper. Specifically, it describes the censored observation situation through a probit regression model and uses the maximum versoria criterion to suppress the adverse effects of non-Gaussian impulsive noise on the update of filter coefficients and background noise variance, ultimately achieving high steady-state estimation accuracy at a lower computational cost. Under common assumptions, the stability and steady-state mean square deviation (MSD) of the CR-MVC algorithm are analyzed, and the stability step size conditions and MSD theoretical model are also derived. Simulation experiments under censored observations and non-Gaussian impulsive noise environment are provided, confirming the accuracy of the theoretical MSD model and the superiority of the CR-MVC algorithm over other competing algorithms.