Improving Signal Reconstruction with Matched Filters and Neural Networks for the Tunka-Rex Experiment

Tunka-Rex is an antenna array for the detection of radio emissions from extensive air showers generated by ultra–high energy cosmic rays. This emission has a broadband spectrum, which corresponds to pulses with durations of tens of nanoseconds and is measured in the band of 30 to 80 MHz. Matched filtering and artificial neural networks are used to improve signal processing at the Tunka-Rex facility. Matched filtering allows more accurate determination the signal peak time, but the best performance can only be achieved with white noise. Convolutional neural networks with autoencoder architecture are used to improve recognition of noise features in traces. These are implemented in Tunka-Rex signal processing and their performance is compared to that of standard means.