基于自适应稀疏表示的光谱去噪和基线校正  

Research on spectral pretreatment based on adaptive sparse representation

作　　者：朱超[1] 鲁昌华[1,2] 杨凯[1] 陈晓婷[1] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009 [2]中国科学院安徽光学精密机械研究所,安徽合肥230031

出　　处：《微型机与应用》2013年第9期54-56,59,共4页Microcomputer & Its Applications

基　　金：国家"863"重点资助项目(2007AA061504;2009AA063006)

摘　　要：通过分析光谱信号特征,结合稀疏表示理论,提出了一种自适应稀疏表示的光谱去噪方法。该方法对信号分段构造学习样本,分别用OMP法和K-SVD法初始化和过训练原子库。将光谱信号在新的原子库上进行自适应稀疏分解,实现光谱信号去噪。利用信噪比(SNR)、均方根误差(RMSE)、波形相似度(NCC)、峰值平均相对误差(AREPV)四个指标来评价去噪效果。仿真实验结果表明,与小波软阈值和小波硬阈值方法相比,该方法能更好地同时消除噪声和基线漂移。According to the characteristics of the spectrum, combining with the sparse representation theory, a denoising method based on adaptive sparse representation is proposed in this paper. This method deride signal into several segments to construct a learning sample, initialize and train atomic library by OMP and K-SVD method. Realize spectrum denoising via adaptive sparse representation the spectrum based on the new dictionary. This paper use four indicators, including signal-to-noise ratio （SNR）, the root mean squared error （RMSE）, the waveform similarity （NCC） and the peak average relative error （AREPV） to evaluate the denoising effect. The simulation results show that： compared with soft and hard threshold method, method based on sparse representation can better eliminate noise and baseline drift at the same time.

关 键 词：光谱预处理 自适应稀疏表示 去噪 基线校正 

分 类 号：F407.63[经济管理—产业经济]