一种改进的奇异值降噪阶次选取方法用于紫外光谱信号去噪的研究

发布时间：2018-01-24 15:20

  本文关键词： 光谱去噪 奇异值分解 模糊C均值聚类　出处：《光谱学与光谱分析》2016年07期 　论文类型：期刊论文

【摘要】：光谱去噪是光谱检测的重要环节。针对光谱信号易受光谱仪热噪声、现场机械振动以及随机噪声等因素影响,而在线监测系统要求减少人为参数选择对去噪效果的影响,提出利用奇异值分解(SVD)理论对光谱信号去噪。提出一种改进的降噪阶次选取方法:指定奇异值差分谱最大峰值点θ_1为所选阶次下界;利用奇异值、奇异值差分谱综合信息选取阶次上界θ_2;将区间θ_1~θ_2定义为模糊区域,通过模糊C均值聚类求取隶属度,赋予模糊区域内奇异值相应的权重系数。用所提方法对不同信噪比下SO_2紫外光谱信号去噪,将信噪比、均方根误差、波形相似系数、平滑度指标用于去噪效果的评价。去噪结果表明:所提方法完全基于数据驱动,具有较好的去噪效果,能够真实的恢复原始信号。
[Abstract]:Spectral denoising is an important part of spectral detection. Spectral signals are easily affected by thermal noise, field mechanical vibration and random noise. And the on-line monitoring system needs to reduce the effect of human parameter selection on the denoising effect. In this paper, the singular value decomposition (SVD) theory is used to de-noise the spectral signal. An improved method for selecting the order of noise reduction is proposed. The maximum peak value 胃 _ 1 of the singular value difference spectrum is specified as the lower bound of the selected order. Using singular value and singular value difference spectrum synthesis information, the order upper bound 胃 s is chosen. This paper defines the interval 胃 _ s _ 1 ~ 胃 _ s _ 2 as a fuzzy region, and obtains the membership degree by means of fuzzy C-means clustering. The proposed method is used to Denoise the SO_2 ultraviolet spectrum signal under different SNR, and the SNR, root mean square error and waveform similarity coefficient are obtained. The smoothness index is used to evaluate the denoising effect. The denoising results show that the proposed method is completely data-driven, has a good denoising effect and can restore the original signal.
【作者单位】： 武汉大学电子信息学院;广西电力科学研究院;
【基金】：国家自然科学基金项目(50677047) 中国南方电网科技项目(K-GX2011-019) 湖北省科学条件专项基金项目(2013BEC010)资助
【分类号】：O657.32
【正文快照】： 组合电器设备(GIS)内SO气体含量,达到了对设备局部放弓|言 电故障的预警,相关研究见文献[7]。针对紫外光谱信号会受光谱仪热噪声、现场机械振动以及随机噪声等因素影响,光谱技术常用于痕量气体的定性、定量检测m。然而在提出利用Saviteky-Goky滤波对光谱信号降噪,取得了不错现，

本文编号：1460353