摘要
通常的线性滤波技术不能很好地解决非线性时间序列去噪问题,而现有的非线性去噪技术的参数选择往往要依赖于直觉和经验.为此,提出基于互信息判据的小波去噪方法,利用小波进行非线性去噪处理,并以互信息作为去噪处理截止的判定条件,给出了小波去噪算法,分析了其优势,并进行了仿真实验.仿真结果表明,小波算法能更好地处理不平稳和突发的噪声;互信息所指示的优化截断尺度,既较好地保留了信号的动力结构,又有效地实现了非线性噪声过滤.
Linear filtering methods often fail in the case of nonlinear time series disturbed by noise. There are a few nonlinear denoise methods, which make use of locally geometrical structure of reconstructed space to identify noise. But reasonable selection of local radius is often an open problem. A novel wavelet denoise method based on mutual information criterion to determine truncation scales of wavelet bases is proposed. Comparing with Fourier denoise methods, the compact support of wavelet bases could deal with non-stationary and burst noise well. Simulation demonstrates that, by means of mutual information criterion, wavelet denoise could distinguish between dynamical structure of signal and noise effectively.
出处
《天津师范大学学报（自然科学版）》
CAS
2006年第4期73-76,共4页
Journal of Tianjin Normal University：Natural Science Edition
基金
天津市科技发展计划项目(04310941R)
天津市应用基础研究计划项目(05YFJMJC11700)
关键词
非线性时间序列
小波去噪
互信息
nonlinear time series
wavelet denoise
mutual information
