High Performance Matched Filtering Algorithm Based on CUDA

Matched filtering is often used to detect the transmitted pulses in the noisy received signal, which is essential for the digital signal processing in real time. This paper proposed a parallel matched filtering technique, used Compute Unified Device Architecture (CUDA) to optimize large-scale data processing, implemented on NVIDIA GPUs, and improved the computational efficiency of algorithm. Experiment results show that the parallel method is faster than the CPU version, or other traditional computing for the real-time signal detection in the practical applications.