多重谱峰的分解方法  

Decomposition Method of Overlapping Peaks for Nuclear Spectrum

作　　者：黄洪全[1,2] 何子述[1] 方方[2] 龚迪琛[2] 丁卫撑[2] 

机构地区：[1]电子科技大学电子工程学院,四川成都611731 [2]成都理工大学地学核技术四川省重点实验室,四川成都610059

出　　处：《原子能科学技术》2010年第9期1114-1119,共6页Atomic Energy Science and Technology

基　　金：国家教育部博士点基金资助项目(20060616010)

摘　　要：利用高斯混合模型(GMM)良好的逼近能力对能谱进行精确描述,并从统计学角度采用期望最大化(EM)迭代算法以4个重叠的伽马谱峰为例,进行了重叠峰的分解。该方法可对由更多谱峰叠加的重叠峰进行分解,并能保证较小的统计误差,对放射性核素的定量和定性分析具有较大参考价值。The energy spectrum was accurately described by Gaussian mixture model which has good approximation capability,and from statistical point of view,by using expectation maximization method,for instance four severely overlapping gamma peaks were decomposed.The peak of more overlapping peaks can be decomposed,and the statistical error is less.It has great reference value for quantitative or qualitative analysis of radio-nuclide.

关 键 词：高斯混合模型 期望最大化 重叠峰分解 

分 类 号：O657.62[理学—分析化学]