Analysis of skin influence in identification of heroin using singular value decomposition

In this paper, the influence of skin in energy-dispersive X-ray diffraction (EDXRD) spectrum of heroin was studied using singular value decomposition (SVD). The spectra of pure heroin, skin and heroin covered by skin were organized as matrices for SVD after truncation and smoothing. It was demonstrated that the two largest singular values and their corresponding left and right singular vectors of each matrix could reconstruct the matrix in the permissible error and contained enough information of the matrix. We extracted the two largest singular values of each matrix as two dimensions of the feature point of the corresponding spectrum. The feature points of different samples were clustered and a linear relationship was proved to be between and movement of feature point and thickness of component of skin, such as fat and muscle. This indicated that the method of SVD may be suitable for identification of heroin covered by skin.