Wang Xin, Li Yan, Wei Hao-yun, Ren Li-bing, Qi Yang
Guang Pu Xue Yu Guang Pu Fen Xi. 2015 May;35(5):1199-202.
The Classical Least Square regression (CLS) is one of the most popular regression methods in FTIR quantitative estimation. However, CLS is the best unbiased estimator only under the assumption that error (noise) in the spectrum has equal variance, which usually is not the case in FTIR. This paper proposed a noise calibration method for FTIR spectrum analysis. Based on measured variance of noise in the FTIR spectrum by computer, the Weighted Least Square regression (WLS) method is used in quantitative estimation. The experiment results showed that the WLS performs much better than CLS in quantitative estimation of VOCs pollution.
经典最小二乘回归（CLS）是傅里叶变换红外光谱（FTIR）定量估算中最常用的回归方法之一。然而，CLS只有在光谱误差（噪声）具有等方差的假设下才是最佳无偏估计量，而在FTIR中通常并非如此。本文提出了一种用于FTIR光谱分析的噪声校准方法。基于计算机测量的FTIR光谱噪声方差，在定量估算中采用加权最小二乘回归（WLS）方法。实验结果表明，在挥发性有机化合物（VOCs）污染的定量估算中，WLS的性能比CLS要好得多。
