Hybrid Fuzzy-Genetic Approach Integrating Peak Identification and Spectrum Fitting for Complex Gamma-Ray Spectra Analysis

A novel hybrid approach for analysis of complex gamma-ray spectra of various origins is described and the test results using spectra obtained from a sodium iodide detector (NaI) are presented. This novel approach exploits the synergism of two artificial intelligence tools; fuzzy logic and genetic algorithms, where the two are merged to identify isotopes and their respective contribution in a given spectrum. The fuzzy logic module focuses on identifying isotopes in the spectrum, while the genetic algorithm (GA) fits and subsequently computes the fractional abundances of the identified isotopes. The fitting of the spectrum is controlled by an assessment procedure based on the test for significance of abundance coefficients, and on the computation of Theil coefficients. This unique synergism between fuzzy logic and GA presents a novel mechanism for automated selection of isotopes for use in spectrum fitting, and as a result eliminates manually-based fitting and/or user intervention. A variety of test cases-including NaI real measured spectra-are used to benchmark this new approach. In addition, the performance of the hybrid method is compared to the multiple linear regression (MLR) fitting approach, along with the combination of fuzzy logic with MLR. This comparison demonstrates a slight superiority of this novel approach regarding accuracy, precision and number of reported false detections.