多重谱峰的分解方法

Decomposition Method of Overlapping Peaks for Nuclear Spectrum

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。

Abstract: 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.