论文部分内容阅读
重叠三维荧光光谱的解析是荧光光谱解析中的难点,非负矩阵分解(NMF)作为一种有效的盲分离方法,能够提取光谱的局部特征和内在联系,克服光谱严重重叠带来的干扰,在解析重叠光谱中具有不可比拟的优势。首先用模拟三维荧光光谱验证了NMF在三维荧光光谱解析中的有效性,然后将四种不同的NMF算法(乘性迭代算法、交替最小二乘算法、二阶方法、投影梯度算法)用于实测的酚类化合物(百里酚、间甲酚、苯酚)三维荧光光谱的解析中,并讨论了在酚类化合物分离应用中四种NMF算法的收敛速度和计算复杂度的差异。实验结果表明,四种方式的NMF标准偏差均在0.06%以下,其中交替最小二乘算法在收敛行为和鲁棒性上最为优越。
The analysis of overlapping 3D fluorescence spectra is a difficult point in the analysis of fluorescence spectra. Non-negative matrix factorization (NMF), as an effective blind separation method, can extract the local features and internal relations of spectra and overcome the interference caused by serious overlapping of spectra. Analysis of overlapping spectra has an incomparable advantage. First of all, the validity of NMF in three-dimensional fluorescence spectroscopy was verified by simulated three-dimensional fluorescence spectroscopy. Then, four different NMF algorithms (multiplicative iterative algorithm, alternating least squares algorithm, second-order method, projection gradient algorithm) Phenolic compounds (thymol, m-cresol, phenol) three-dimensional fluorescence spectra of the analysis, and discusses the application of the separation of phenolic compounds four kinds of NMF algorithm convergence rate and the complexity of the difference. The experimental results show that the NMF standard deviations of the four methods are below 0.06%, of which the ALM algorithm is the most superior in convergence behavior and robustness.