噪音(视频)
数学
非线性系统
噪声测量
控制理论(社会学)
噪声功率
系统标识
高斯噪声
均方误差
算法
公制(单位)
应用数学
统计
计算机科学
功率(物理)
降噪
数据建模
人工智能
工程类
图像(数学)
数据库
物理
量子力学
运营管理
控制(管理)
作者
Binwei Weng,Kenneth E. Barner
标识
DOI:10.1109/tsp.2005.849213
摘要
Nonlinear system identification has been studied under the assumption that the noise has finite second and higher order statistics. In many practical applications, impulsive measurement noise severely weakens the effectiveness of conventional methods. In this paper, /spl alpha/-stable noise is used as a noise model. In such case, the minimum mean square error (MMSE) criterion is no longer an appropriate metric for estimation error due to the lack of finite second-order statistics of the noise. Therefore, we adopt minimum dispersion criterion, which in turn leads to the adaptive least mean pth power (LMP) algorithm. It is shown that the LMP algorithm under the /spl alpha/-stable noise model converges as long as the step size satisfies certain conditions. The effect of p on the performance is also investigated. Compared with conventional methods, the proposed method is more robust to impulsive noise and has better performance.
科研通智能强力驱动
Strongly Powered by AbleSci AI