Automatic pattern recognition on electrical signals applied to neutron gamma discrimination

The electrical pattern recognition can be useful in several applications; generally it is used to detect particular events or anomalies in the signal under analysis or to identify precursors, especially in electrophysiology. Each application requires customized algorithms and appropriate signal processing capabilities. In this article we present the pattern recognition applied to neutron and gamma scintillator analysis; the algorithm can be used considering that the incident particles on the detector produce pulses having different shape. The discrimination of particles is performed starting from a reference patterns set. The algorithm has been designed to be efficiently implemented in programmable logic gate array; anyway, considering the broadband of the signals under analysis, the real time implementation needs simply reference set based on a limited number of patterns due to technological constrains. The algorithm can also be applied off line by using a more complex reference pattern sets in order to detect and to classify the pile-up event, or to compress the scintillator data. The proposed pattern recognition algorithm is based on the cross-correlation operator and on the Euclidean distance between the reference pattern and the shape of the signal under analysis. The automatic pattern recognition algorithm and its simulations are reported in the article. In order to verify the performances in the case of scintillator signals, the algorithm has been applied on data acquired by two scintillator systems irradiated by a neutron-γ source at the Frascati Tokamak Upgrade laboratories. The results confirm the suitability of the method and its future usability. With minor changes the systems can be used in different diagnostic fields.