基于小波和小波包变换的苯丙酮尿症傅里叶变换衰减全反射红外光谱筛查模型的研究  

A Phenylketonuria Screening Model Based on FTIR/ATR Spectra,Wavelet and Wavelet Packet Transform

作　　者：蔡嘉华 尹嵩杰 陈超[1,2,3] 王淑美[1,2,3] 梁生旺[1,2,3] 

机构地区：[1]广东药学院中药学院,广东广州510006 [2]国家中医药管理局中药数字化质量评价技术重点研究室,广东广州510006 [3]广东高校中药质量工程技术研究中心,广东广州510006

出　　处：《分析科学学报》2016年第4期458-462,共5页Journal of Analytical Science

基　　金：国家自然科学基金(No.81274059);广东省自然科学基金(No.S2012010009166);广州市珠江科技新星基金(No.2014J2200021);广东省教育厅优秀青年教师基金(No.Yq2013102)

摘　　要：应用小波和小波包变换对傅里叶变换衰减全反射红外光谱(FTIR/ATR)进行去噪处理,以提高苯丙酮尿症(PKU)筛查模型的性能。首先优化小波和小波包变换的参数,然后分别对原始光谱(OS)、9点平滑光谱(9S)和一阶微分9点平滑光谱(1D9S)进行去噪处理,以均方根误差(RMSE)、平均相对误差(MRE)、预测准确率(Acc)等为指标,考察小波和小波包变换对模型性能的影响。结果与变换前相比,模型性能均有所提高,其中小波变换以1D9S+sym12处理结果为最优,而小波包变换以1D9S+sym1为最优;Acc全部提高为100%。In this paper, wavelet and wavelet packet decomposition methods were used to denoise the FTIR/ATR spectra,which were applied to improve the predictive capability of phenylketonuria（PKU） screening models. The parameters of wavelet and wavelet packet decomposition were firstly optimized. And raw spectra（OS）,9-point smoothing spectra（gS） as well as 9-point smoothing coupled with first differential spectra（1DgS） were then denoised. The predictive capabilities of constructed models based on the above-mentioned spectra were evaluated by root mean square error （RMSE）, mean relative error （MRE）,predictive accuracy（Acc）,etc. The results showed that the predictive accuracies of the three models were improved compared with those before wavelet transformation, in which 1D9S spectra with sym12 function was the optimal one. For wavelet packet decomposition,the optimal one was 1D9S with syml function. All values of Acc were increased to 100%.

关 键 词：小波变换 小波包变换 多模型共识 苯丙酮尿症 傅里叶变换衰减全反射红外光谱 

分 类 号：O657.33[理学—分析化学]