Feature extraction in fingerprint of medicinal herbs via EMD and fractal technique

The wavelet fractal features of fingerprint of medicinal herbs based on wavelet transform has anti-jamming property against the change of sample concentration. However, the recognition rate based on wavelet fractal features is not satisfied when fingerprint of medicinal herbs has some slight changes of concentrations, the number of peak and peak drift of sample are processed in specially situation. In this paper, we put forward the conception of fractal feature of chromatographic profile of medicinal herbs using empirical mode decomposition (EMD) to obtain intrinsic mode functions (IMF) from high to low frequency, and establish a novel method to extract fractal feature of chromatographic profile of medicinal herbs. We obtain some sub modes through the chromatographic profile of medicinal herbs by EMD, and then by using fractal theory, we achieve the fractal dimensions of each mode. We apply it to the recognition of the fingerprints of glycyrrhiza. Experiments show that EMD fractal features have better recognition rate than that of the traditional ones in case of the concentration-change, i.e. the number of peak and peak drift of sample have slight changes.