基于经验模态分解与回归分析的空间外差光谱目标提取

为识别空间外差光谱仪探测目标干涉信号的特征信息，提出一种基于经验模态分解与回归分析的空间干涉谱目标提取方法。首先对预处理后的光谱进行经验模态自适应分解，得到各阶次固有模态分量并分别计算它们与原始光谱信号的Pearson相关系数，根据相关系数分选准则判定背景与目标信息重构的分界点。然后计算重构背景与实测背景间的Pearson相关系数来判定经验模态分解结果。对信号主导的固有模态分量利用小波软阈值进行消噪，重构较纯净的目标特征信息；利用目标特征信息与原始干涉光谱信息进行多元线性回归分析获得最佳的近似滤波系数，构造滤波器并应用到目标信号，提取目标。最后通过差谱信号与提取的目标光谱的Pearson相关系数来判别提取的目标信号。实验结果表明:经验模态分解可将背景与目标近似分离；在未知背景信号情况下，利用经验模态分解与回归分析可实现钾共振双线特征光谱的提取。

Extraction of spatial heterodyne spectroscopy target based on empirical mode decomposition and regression analysis

The algorithm was proposed based on the empirical mode decomposition and regression analysis to extract and identify the characteristic information of spatial heterodyne spectroscopy. The spectrum which was obtained by pre-processing the original probe data was decomposed into several intrinsic mode function components by empirical mode decomposition and the each order IMF's Pearson correlation coefficient was calculated with the original spectral signal. According to the correlation coefficient classification criteria, the demarcation point of the background and target information reconstruction will be determined. Then the Pearson correlation coefficient between the reconstructed background and the measured background was calculated to determine the empirical mode decomposition results. At the same time, the signal-dominated components were de-noised respectively by the wavelet soft threshold and then the pure target characteristic signal was reconstructed. By using multiple linear regression analysis to process the target characteristic information and the original interference spectral information, the optimal coefficients of time-domain filtering will be obtained. The filter will be constructed to extract the target. Finally, the signal of extracted target will be identified by Pearson correlation coefficients. The experimental results show that the background and the target can be separated by the empirical mode decomposition. In the case of unknown background signal, the empirical mode decomposition and regression analysis can be used to extract the characteristic spectrum of potassium resonance.