基于小波变换与分形的中药指纹图谱的特征提取  

A Combination of Fractal and Wavelet for Featuer Extraction of the Chromatographic Profile of Medicinal Herbs

作　　者：杜建卫[1] 程乾生[2] 刘乃强 林立[2] 

机构地区：[1]北京石油化工学院数理部,北京102617 [2]北京大学数学科学学院,北京100871 [3]河南药品检验所

出　　处：《信号处理》2006年第5期719-723,共5页Journal of Signal Processing

摘　　要：探讨基于小波变换与分形的中药指纹图谱特征提取方法。使用小波变换将中药指纹图谱分解至不同的分辨尺度,然后利用分形理论计算各尺度下的分量的分形维数,用分形维数参量代替指纹图谱的采样值,对中药进行分析识别。试验结果表明,使用小波基分形参量对一药物指纹图谱的实验条件的辨别效果要优于单纯使用分形维数和使用指纹图谱的采样值。Discuss the application of the fractal expression way based on wavelet analysis for feature extraction of the chromatographic profile of medicinal herbs. In this paper, the chromatographic profile of medicinal herbs is decomposed into components at different resolution levels using wavelet analysis, then the fractal dimensions of these components are computed as the chemical fingerprints. The chromatographic fingerprint is characterized by the vector composed of these chemical fingerprints, which can represent the chemical patterns of different categories of complex samples. Simulations reveal that the recognition result obtained by the fractal fingerprints of wavelet bases is better than that using only the original chromatographic profile data or the fractal dimension of the original chromatographic profile data.

关 键 词：中药指纹图谱 小波变换 分形维数 

分 类 号：O18[理学—数学]