Light charged particle classification using subspace identification methods and neural networks

The problem considered in this work is the classification of the particles produced by the collision of a heavy ion beam over a target. Each particle is captured by a detector, and results in a signal (which is the impulse response of a dynamic linear system), which is measured by a digital acquisition system. The assumption made herein is that the shape of the impulse response contains complete information on the particle, and the classification can be done by pulse-shape analysis. The signature on which the classification is performed is the atomic mass number (A) and the atomic charge (Z). In this work, a complete procedure for the particle identification is proposed. The main idea is to use the cascade of a state-space identification algorithm and a parametric nonlinear map using the model parameters as input regressors. The algorithm has been tested on a large set of impulse-responses, and provides a fully-automatic accurate classification of the isotopes. This work focuses on light charged particles (LCP), which are the most difficult to detect and to classify. All the experiments are made with the large detector array CHIMERA (charge heavy ions mass and energy resolving array)