摘要
将离散小波变换、小波包变换、傅里叶变换和离散余弦变换与主组分回归方法结合构成4种离散变换主组分回归方法,编制了离散变换主组分回归方法的计算程序。将离散变换主组分回归方法用于处理对硝基甲苯、对硝基酚和对硝基苯胺混合物的重叠紫外吸收光谱数据。结果表明,离散变换主组分回归方法优于主组分回归方法,试样质量浓度的预测值与实际值的相对预测标准误差由3.81%降至约1.11%。
Four kinds of discrete transform principal component regression methods were constructed by combining wavelet packet transform （WPT）, wavelet transform （WT）, Fourier transform （FT） and discrete cosine transform （DCT） with principal component regression, and calculating programs were designed to perform calculations of discrete transform principal component regression methods. The methods were used to resolve overlapping spectra of the mixtures of p-nitrotoluene, p-nitrophenol and p-nitroaniline. The experimental results indicated that discrete transform principal component regression methods were better than principal component regression method and the predictive relative standard errors between the actual and estimated values of mass concentration of samples decreased from 3.81% to about 1.11%.
出处
《分析科学学报》
CAS
CSCD
2007年第5期579-582,共4页
Journal of Analytical Science
基金
国家自然科学基金(No.20667002)
内蒙古自然科学基金(No.200408020210)
关键词
离散变换
主组分回归
吸收光谱
重叠峰
Discrete transforms
Principal component regression
Overlapping peaks
