基于CUDA的并行调制识别算法研究-电子与通信工程专业论文.docxVIP

万方数据 万方数据 摘 要 摘 要 信号的调制识别技术是信号分析领域一项不可或缺的关键技术，在电子战中 具有重要的战术与战略意义。以 CUDA（Compute Unified Device Architecture，计 算统一设备架构）为代表的 GPU 通用计算技术的普及开辟了新的技术途径。本文 重点研究了 CUDA 架构下调制识别算法的并行化设计。 本文首先介绍了基于统计矩的通信调制识别算法和基于短时傅里叶变换的雷 达脉内调制识别算法，给出了具体的算法流程和仿真验证。而后介绍了 CUDA 的 编程模型、软件体系和存储器模型，以及 CUDA 程序的性能评估策略和优化策略， 经过对调制识别算法的并行性分析，给出了调制识别计算任务的分解过程及程序 流程。最后基于 CUDA 平台实现了基于统计矩的通信调制识别算法和基于短时傅 里叶变换的雷达脉内调制识别算法的并行化，并结合仿真结果对两类并行调制识 别算法进行了性能分析。实验结果表明，相比基于 CPU 的串行算法，基于 CUDA 的并行调制识别算法能够取得较好的加速比。 关键词：CUDA 并行化 调制识别 统计矩 短时傅里叶变换 Abstract Abstract Signal modulation recognition technology which is an essential key technology in the field of signal analysis has an important tactical and strategic significance in the electronic warfare. The popularity of general purpose GPU technology represented by CUDA (Compute Unified Device Architecture) opens another approach. This paper focuses on the realization of parallel modulation recognition algorithm within the CUDA architecture. At first, the communication modulation recognition algorithm based on statistical moment and radar intra-pulse modulation characteristic recognition algorithm based on Short Time Fourier Transform are introduced, the detailed algorithm process and the simulation are given. Next, the CUDA programming model, software architecture and memory model are introduced, and the performance evaluation strategy and optimization strategies are described. The decomposition process of computational tasks and the parallel algorithm process are given after analyzing the parallelism of modulation recognition algorithm. At last, the parallelization of modulation recognition algorithm based on CUDA platform is realized,and the performance analysis of the two modulation recognition algorithms are given combining the simulation results. The experimental results show that the parallel modulation recognition algorithm based on CUDA can obtain a good speed-up compared with the serial algorithm based on CPU. Keywords: CUDA, Parallelization, Modulation recognition,