Tensor Decomposition Techniques for Mixed Radiation Signal Separation: A Comparative Study
Imane Ahnouz, Hanan Arahmane, Rajaa Sebihi · 2025
Accurate discrimination between neutrons and gamma rays in mixed radiation fields is essential for effective neutron detection, which plays a critical role in applications such as homeland security and medical treatments. Various traditional and advanced techniques have been proposed to enhance the performance and precision of neutron-gamma discrimination. This study provides a comparative analysis of three tensor decomposition-based methods—Multiway Blind Source Separation (MBSS), Sequentially Truncated Multilinear Singular Value Decomposition (ST-MLSVD), and Nonnegative Tensor Factorization (NTF)—for addressing this challenge. The results show that MBSS and ST-MLSVD significantly outperform NTF in terms of signal isolation, as indicated by the mean Signal-to-Interference Ratio values. Additionally, all three methods effectively characterized the sources, with correlation coefficient calculations further validating their performance in distinguishing neutron and gamma signals.