A method for discriminating neutron and gamma waveforms based on a comparison of differences between pulse feature heights

Accurate discrimination of neutron and gamma signals is essential for precise neutron detection. However, traditional discrimination methods face challenges under less-than-ideal conditions, such as varying signal-to-noise ratios, varying sampling rates, and pulse pile-up, leading to decreased accuracy. In this study, a simpler machine learning-based method was proposed for discriminating thermal neutron/gamma pulse shapes. The method was tested on data from a CLYC detector using a ²⁴¹Am-Be neutron source. A comparison was made with three traditional methods and three other machine learning-based methods. The proposed method exhibited excellent anti-noise capabilities, particularly at low signal-to-noise ratios. Even with a standard deviation of noise reaching 0.05, the proposed method achieved an accuracy of 90%, surpassing the performance of the six discrimination methods evaluated.