中药色谱指纹图谱的小波变换及分形表达方法

提出一种基于小波变换的色谱指纹图谱分形表达方法。该法运用小波变换将色谱谱线分解至不同分辨尺度，然后计算各尺度分量的分形维数，用色谱的小波基分形参量替代色谱指纹图谱的采样值。仿真实验结果表明，色谱小波基分形参量对谱峰保留时间漂移具有较好的抗干扰能力。以当归和川芎两种中药材的品种及道地性鉴别分类问题为实例，比较研究色谱采样值与小波基分形参量，k-近邻法的交叉验证计算结果表明，小波基分形参量的分类效果优于色谱采样值。

A Method for Characterizing Chromatographic Fingerprint by Integrating Wavelet Transform with Fractal Dimension

A method for characterizing chromatographic fingerprint by integrating wavelet transform with fractal dimension was proposed. The chromatographic profile is decomposed into the components of different resolution levels by wavelet transform, and the fractal dimensions of these components are then computed. As a result, chromatographic signal data are replaced by the wavelet-based fractal parameters for characterizing chromatographic fingerprint. It has been substantiated by the simulation experiment that, for the disturbance caused by variation of retention times of peaks, the anti-disturbing performance of the wavelet-based fractal parameters is better than that of chromatographic signal data. A typical example of identifying the species and habitats of two herbs, i.e. Ligusticum Chuangxiong and Angelica Sinensis (Oliv.) Diels, was used for comparative analysis between the wavelet-based fractal parameters and chromatographic signal data. The cross-validation results of k Nearest Neighbor showed that the classification accuracy obtained with the wavelet-based fractal parameters was greater than that obtained with chromatographic signal data.