Speaker
Description
Radio detection is widely used to instrument large volumes of ice in searches for ultra-high-energy neutrinos, but its sensitivity is strongly affected by the trigger threshold near the thermal-noise floor. Conventional threshold-based triggers often become limited in the weak Askaryan-like impulse regime near the noise floor, where thermal noise can overwhelm signal events, significantly reducing detection sensitivity. We propose and implement a convolutional neural network trigger that classifies continuous full-band input waveforms while maintaining sub-watt power consumption, enabling effective retention of weak signal-like impulses. The implementation results demonstrated that this novel trigger has the potential to reduce the effective trigger threshold from the conventional 4V_{\mathrm{RMS}} level toward 2V_{\mathrm{RMS}}, with a projected 5–10-fold improvement in the detection rate in the near-threshold regime.