Real-time signal processing for FM-based passive bistatic radar Using GPUs

This paper addresses the problem of the real-time signal processing work of a FM-based passive bistatic radar (PBR) system. The signal processing architecture is fully deployed on Graphic Processing Units (GPUs) using Compute Unified Device Architecture (CUDA). By exploring the parallelism of the algorithms, we manage to assign the data processing task from two carrier frequencies on one NVIDIA Tesla C2075. Numbers of mapping strategies are explored during the research and are presented in this paper. While yielding a substantial performance on real-time signal processing, the GPU implementation delivers speedups of 37x over standard Central Processing Units (CPUs). Besides, the GPU implementation exhibits flexibility in sustaining data load from more carrier frequencies.