基于Storm的分布式实时信号处理系统

发布时间：2018-02-23 13:18

本文关键词： 信号处理系统 Storm 实时并行计算波束形成　出处：《计算机应用》2017年S1期 　论文类型：期刊论文

【摘要】：针对传统基于数字信号处理器(DSP)的实时信号处理系统存在的编程复杂、硬软件耦合紧、可靠性和扩展性差等问题,提出了一种基于大数据流式实时处理系统Storm的新型信号处理机架构。该架构利用通用处理器和Storm系统替代DSP处理器和其相应的数据流程序框架,对阵列信号处理算法进行模块化分解,并揉入Storm拓扑中,由Storm通用处理器集群完成数据的传输和处理。利用常规波束形成算法对新型系统进行测试,其对一拍数据进行傅里叶变换与波束形成计算的时间分别为59.13 ms和5.27 ms,远小于两个数据节拍之间的时间间隔。利用人工去除节点来模拟节点失效,系统能够进行自我修复并继续运行。结果与分析表明新型架构完全能够满足水声阵列信号处理系统对实时性的要求,且可靠性高、可扩展性与程序可复用性好。
[Abstract]:Aiming at the problems of complex programming, tight coupling of hardware and software, poor reliability and expansibility of traditional real-time signal processing system based on DSP, etc. A new signal processor architecture based on large data stream real-time processing system (Storm) is proposed, in which the DSP processor and the corresponding data stream program framework are replaced by the general purpose processor and the Storm system. The array signal processing algorithm is modularized decomposed and rubbed into the Storm topology, and the data is transmitted and processed by the Storm general processor cluster. The conventional beamforming algorithm is used to test the new system. The time of Fourier transform and beamforming is 59.13 Ms and 5.27 Ms, respectively, which is much smaller than the interval between the two data beats. Results and analysis show that the new architecture can fully meet the real-time requirements of underwater acoustic array signal processing system, and has high reliability, good scalability and program reusability.
【作者单位】：声场声信息国家重点实验室(中国科学院声学研究所);中国科学院大学电子电气与通信工程学院;
【基金】：国家自然科学基金资助项目(61571436)
【分类号】：TN911.7;TP332

【参考文献】