Real-time Pulse Search and Positioning Method for Baseband Data Based on Matched Filtering

Fast radio bursts (FRBs) are the most intense radio explosions known, yet their origins remain a significant astrophysical mystery. High-time-resolution pulse searching has opened new avenues for studying narrow and faint pulses, promising substantial improvements in pulse detection efficiency and accuracy and potentially leading to significant scientific discoveries. However, the increase in time resolution results in exponential data volume growth, presenting considerable challenges for data processing. The real-time processing of vast amounts of data and the rapid detection of FRB pulse signals have become bottlenecks in high-time-resolution pulse search research. To address these challenges, we propose a real-time pulse search and positioning method for baseband data based on matched filtering. This method performs preliminary pulse searching and positioning within the baseband data, enabling the selection of effective data segments for subsequent dedispersion and analysis, thereby significantly reducing the data-processing burden. Compared to traditional coherent dedispersion methods, our approach exhibits superior performance in both pulse detection efficiency and dispersion tolerance. Utilizing GPU parallel acceleration, the method can process baseband data streams with 1 ns time resolution in real time, facilitating the detection of narrower and fainter pulse signals. Analysis of 1 ns time-resolution baseband data from FRB 121102 not only successfully identifies all known pulses, but also reveals three new pulses, demonstrating the feasibility and effectiveness of our method.