Aiming at the problem of online complex Raman spectra signal with fluorescence background and serious peak overlapping, a new method is provided to extract and separate useful signals from fluorescent background signals. Firstly, wavelet transform is used to deduct the fluorescent background ( i. e. low frequency components) in original signal. Then immune algorithm is used to separate useful signals. Because immune algorithm may have slow convergence rate and even have non-convergence in multi-component mixture, the standard signals of immune algorithm is weighted by cumulative percentage which is obtained from principal component analysis and overlapping Raman spectra is recognized correctly and quickly. It also meets the requirements of accurate and rapid peak recognition when online Raman spectrometer is used in industry field.%针对在线拉曼光谱信号存在荧光背景和严重谱峰重叠的情况,提出了一种从含荧光背景信号中提取有用信号并对其进行充分分离的谱峰识别方法.首先采用小波变换扣除原始信号中的荧光背景(即低频成分),然后采用免疫算法对有用信号进行分离.由于免疫算法在组分较多情况下收敛速度慢甚至不收敛,因此,又采用主成分分析(PCA)法得到累计百分比,对免疫算法中的标准信号进行加权,从而完成对重叠拉曼光谱信号正确、快速地识别,并很好的满足了在线拉曼光谱仪在工业现场谱峰识别准确快速的要求.
展开▼
