彭思愿2,
陈霸东1,,
1.西安交通大学人工智能与机器人研究所 西安 710049
2.南洋理工大学电气与电子工程学院 新加坡 639798
基金项目:国家自然科学基金-深圳市联合研究项目(U1613219),国家自然科学基金(91648208, 61976175)
详细信息
作者简介:卢明飞:男,1987年生,博士生,研究方向为自适应信号处理、模式识别与机器学习等
彭思愿:男,1991年生,博士生,研究方向为非负矩阵分解、信息论学习和自适应滤波算法等
陈霸东:男,1974年生,教授,研究方向为先进信号处理与脑机接口、机器学习与认知计算以及新型神经网络计算模型
通讯作者:陈霸东 chenbd@mail.xjtu.edu.cn
中图分类号:TN713计量
文章访问数:412
HTML全文浏览量:193
PDF下载量:82
被引次数:0
出版历程
收稿日期:2020-04-21
修回日期:2020-10-21
网络出版日期:2020-11-18
刊出日期:2021-02-23
Convex Combination of Multiple Adaptive Filters under the Maximum Correntropy Criterion
Mingfei LU1,Siyuan PENG2,
Badong CHEN1,,
1. Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University, Xi’an 710049, China
2. School of Electronics and Information Engineering, Nanyang Technological University, 639798, Singapore
Funds:The National Natural Science Foundation-Shenzhen Joint Research Program (U1613219), The National Natural Science Foundation of China (91648208, 61976175)
摘要
摘要:基于最大互相关熵准则(MCC)的自适应滤波算法在非高斯噪声环境下具有强鲁棒性,得到了广泛应用。然而,传统MCC滤波算法在选择参数时依然受到收敛速度与稳态精度之间固有矛盾的困扰。为解决这一问题,该文提出一类多凸组合MCC算法,能够充分发挥不同参数组合下滤波算法的性能优势,从而获得更好的信道跟踪能力。理论分析得出了所提算法的均值收敛条件和稳态均方误差,同时,仿真实验表明所提算法在对抗高斯和非高斯噪声时均具有收敛快、稳态精度高的特点。
关键词:自适应滤波/
信道估计/
最大互相关熵准则/
凸组合
Abstract:The adaptive filtering algorithms under the Maximum Correntropy Criterion (MCC) show strong robustness against impulsive noises. The original MCC adaptive filter, however, still suffers from a compromise between convergence rate and misadjustment when choosing parameters. To address this issue, a convex combination approach is proposed in this paper, where multiple MCC adaptive filters with different step-sizes and kernel widths are combined together to yield fast convergence speed and lower misadjustment. Theoretical analysis on convergence of the new approach demonstrates that it can achieve more desirable performance than the original MCC adaptive filter as well as convex combination of two MCC adaptive filers with different step-sizes or kernel widths. Simulation results confirm the excellent performance of the new method.
Key words:Adaptive filtering/
Channel estimate/
Maximum Correntropy Criterion(MCC)/
Convex combination
PDF全文下载地址:
https://jeit.ac.cn/article/exportPdf?id=2ab1c9fe-8647-4092-95c4-2becf55e3c43