Optimal Signal Selection for a Highly Segmented Detector

This work presents an extensive study of signal detection against noise for a high-energy calorimeter (energy measurement) in the context of particle collider experiments. We aim at selecting the calorimeter cells (10,000 readout channels available, most of them with no signal) that should be considered for energy reconstruction. Several techniques for the signal detection are employed such as Maximum Likelihood, independent component analysis and neural processing. The results show that the neural network approach for signal detection surpasses the other techniques in terms of both performance and implementation complexity.