Novel information theory based method of gamma-ray spectra identification

In this paper, a relative entropy based method is proposed to identify the gamma-ray spectra of radioactive sources. Firstly, Principal Component Analysis (PCA) algorithm is used to compress data and construct an eigenspace of the gamma-ray spectrum. Then, Randomization Technique (RT) is adopted to normalize the eigenvalue of the gamma-ray spectrum in eigenspace. Hence, the eigenspaces of gamma-ray spectra can be regarded as probability spaces. Finally, the relative entropy of two probability spaces is defined to measure the difference between two contrasted gamma-ray spectra. It was experimentally demonstrated that the proposed method could perform better judgment about the identity of two gamma-ray spectra over most existing methods. The proposed method has the characteristics of less calculation and higher robustness for impact factors of statistic fluctuations, peaks drift and background.