净分析信号算法用于近红外模型优化的研究

将净分析信号概念引入多元复杂的银杏叶提取物分析中，利用净分析信号算法对近红外光谱图进行处理，并用净分析算法结合二维相关光谱算法对模型进行了优化，最终建立槲皮素、山柰素和异鼠李素三种组分同时测定的近红外定量模型。利用两种净分析信号算法(Lorber的NAS算法和Goicoechea 与 Olivieri的HLA/GO算法)作为光谱预处理方法，用一系列主因子重构后的光谱图建立校正模型，并通过净分析信号-二维相关光谱算法对图谱中主对角线处的波段进行选择，优化模型建立所采用的波段。所采用的两种基于净分析信号的预处理方法成功的应用于银杏叶中三种黄酮：槲皮素，山奈素和异鼠李素含量的同时测定。净分析信号预处理方法显著降低了建模所需主因子数，提高模型的稳健性。净分析信号-二维相关光谱算法能够优化建模波段，提高模型的预测能力。结论：净分析信号算法能够应用于近红外光谱模型预处理和优化过程，在复杂多组分且干扰未知植物药含量分析的研究中具有实际应用价值。

Study on the Application of NAS-Based Algorithm in the NIR Model Optimization

In this paper, net analysis signal (NAS)-based concept was introduced to the analysis of multi-component Ginkgo biloba leaf extracts. NAS algorithm was utilized for the preprocessing of spectra, and NAS-based two-dimensional correlation analysis was used for the optimization of NIR model building. Simultaneous quantitative models for three flavonol aglycones: quercetin, keampferol and isorhamnetin were established respectively. The NAS vectors calculated using two algorithms introduced from Lorber and Goicoechea and Olivieri (HLA/GO) were applied in the development of calibration models, the reconstructed spectra were used as input of PLS modeling. For the first time, NAS-based two-dimensional correlation spectroscopy was used for wave number selection. The regions appeared in the main diagonal were selected as useful regions for model building. The results implied that two NAS-based preprocessing methods were successfully used for the analysis of quercetin, keampferol and isorhamnetin with a decrease of factor number and an improvement of model robustness. NAS-based algorithm was proven to be a useful tool for the preprocessing of spectra and for optimization of model calibration. The above research showed a practical application value for the NIRS in the analysis of complex multi-component petrochemical medicine with unknown interference.