项目名称: 高效稳健的自适应绝对偏度滤波算法研究
项目编号: No.61271341
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 无线电电子学、电信技术
项目作者: 张家树
作者单位: 西南交通大学
项目金额: 88万元
中文摘要: 虽然利用了三阶统计量的偏度最大化自适应算法在一些应用中比线性自适应算法和基于四阶矩的自适应算法等具有更好的滤波性能,但基于偏度最大化的自适应算法在长拖尾非高斯噪声环境下的性能退化明显,对脉冲干扰尤其敏感。为此,本项目拟采用绝对偏度与稳健统计理论,发展出一系列高效稳健的自适应绝对偏度算法。具体研究内容包括:(1)非高斯环境下的自适应滤波误差建模与稳健参数估计;(2)最小均方绝对偏度算法及其性能分析;(3)仿射投影绝对偏度算法及其性能分析;(4)递归最小二乘绝对偏度算法及其性能分析;(5)共轭梯度绝对偏度算法及其性能分析。 鉴于本项目拟发展的自适应绝对偏度算法将充分利用误差信号的一阶偏差、二阶和三阶绝对统计量和稳健估计理论,预期取得的自适应绝对偏度算法将具有计算复杂性低、收敛速度快、稳态误差小和稳健性等优点,能更好地满足工程应用的需求,将进一步丰富和发展非线性自适应信号处理的理论与方法。
中文关键词: 自适应滤波;非高斯;非平稳;性能分析;鲁棒性
英文摘要: Adaptive algorthms based on skewness maximum have shown better convergence and tracking performance than linear adaptive algorithms and those based on fourth moments. However, adaptive skewness algorthms will significantly degrade their convergent performance in the presence of heavy tailed noise, especially be sensitve to the impulsive noise. Therefore, this project will develop a series of efficient and robust adaptive absolute skewness algorithms from robust statistics and absolute skewness metrics. The research contents include: (1) adaptive filtering error modeling and robust parameter estimation under non-Gaussian environments; (2) least mean absolute skewness algorithms and their performance analysis; (3) affine projection absolute skewness algorithms and their performance analysis; (4) recursive least absolute skewness algorithms and their performance analysis; (5) conjugate gradient absolute skewness algorithms and their performance analysis. These adaptve absolute skewness algorithms will make full use of the absolute deviation, variance and third absolute moments of adaptive filtering error, and their robust estimations, so we can believe that these robust algorithms will be of low complexity, fast convergence, small steady-state misadlignment, numerical stability and robustness against impusilve no
英文关键词: adaptive filtering;non-Gaussian;non-stationary;performance analysis;robust