乙醇含量拉曼光谱检测中，拉曼光谱信号中的各种噪声及光谱荧光造成的基线漂移和样品池背景等，影响了校正模型的预测精度。利用总体平均经验模态分解，将光谱信号分解成若干无模态混叠的内在模式分量，根据排列熵的信号随机性检测判据判断出代表背景信息和噪声信息的内在模式分量，将其置零即可同时消除拉曼光谱中的噪声与背景。将总体平均经验模态分解与排列熵相结合的预处理方法应用于乙醇含量的拉曼光谱检测中，并与小波变换和平均平滑滤波做了对比。实验结果表明：应用总体平均经验模态分解与排列熵相结合的方法能够有效的同时消除乙醇含量拉曼光谱检测中的噪声和背景信息，提高校正模型的预测精度，且使用简便，无需参数设置，对乙醇含量拉曼光谱检测具有实用价值。%In the process of detecting ethanol content by Raman spectra ,the precision of correction model prediction is affected by noise and baseline drift ,which is caused by the spectral fluorescence and sample pool’s background .Use ensemble empirical mode decomposition to decompose spectrum into several intrinsic mode functions ,which are without aliasing .The permutation entropy is employed to judge the intrinsic mode functions .Set the intrinsic mode functions which are on behalf of noise and back-ground to zero ,and then the signal is without noise and background .In this paper combine ensemble empirical mode decomposi-tion and permutation entropy ,and apply to the Raman spectrum ,which are used to detect ethanol content .At the same time compare with wavelet transform and average smoothing filter .The experimental result shows that the application of empirical mode decomposition and permutation entropy can effectively eliminate the noise and background .The precision of correction model prediction is improved .This method simply employs and doesn’t need to set parameters ,which has great value of applica-tion in the process of detecting ethanol content by Raman spectra .
展开▼
