Sigmoid变换的滤波-x四元数最小均方算法
Filtered-x Quaternion Least Mean Square Algorithm of Sigmoid Transformation
张冰妍 1陈晓梅 1钟波2
作者信息
- 1. 华北电力大学 电气与电子工程学院,北京 102206
- 2. 中国计量科学研究院 力学与声学计量科学研究所,北京 100029
- 折叠
摘要
滤波-x最小均方(Filtered-x Least Mean Square,FxLMS)算法是主动噪声控制系统中常用的算法,对中低频噪声有较好的控制作用,但在某些环境噪声中传统的算法可能达不到期望的抑制效果.提出一种基于sigmoid变换的滤波-x四元数最小均方算法,该算法利用四元数的空间特性使噪声信号在超复数域内部相互耦合和关联,并通过sigmoid函数对误差信号进行非线性变换来约束噪声信号以减低对权值更新的影响力度,避免权值在更新过程中发散,从而实现优异的收敛性能以及增强的鲁棒性.同时通过研究步长分析该算法的稳态特性,并在汽车、工厂噪声环境下验证提出算法性能的优越性,仿真结果支持了该结论.
Abstract
FxLMS algorithm is the commonly adopted algorithm in active noise control system,which has a good control effect on medium and low frequency noise,but the traditional algorithms may not reach the desired suppression effect in certain environmental noise.Filtered-x quaternion least mean square algorithm of sigmoid transformation is proposed,which exploits the spatial characteristics of quaternion to make noise signals couple and correlate in the hyper-complex domain.The sigmoid function is applied to perform nonlinear transform on the error signal to constrain the noise signal to avoid the divergence during the weight update progress,so as to achieve better convergence performance and enhanced robustness.The steady state of the proposed algorithm is analyzed,the superiority of the proposed algorithm in noise reduction is verified in volvo and factory noise environments.The simulation results support the conclusions.
关键词
振动与波/主动噪声控制/四元数自适应滤波器/非线性变换/FxLMS/收敛Key words
vibration and wave/active noise control/quaternion adaptive filter/nonlinear transform/FxLMS/convergence引用本文复制引用
基金项目
国家重点研发计划专项(2020YFC2005200)
出版年
2024
国家科技期刊平台
NSTL
万方数据