一种基于Hadoop实现的远场声压预估系统  

A System of Far- field Sound Pressure Prediction Based on Hadoop

作　　者：史书源 程晓斌[1] 王勋[1] 陈艳霞[1] 

机构地区：[1]中国科学院声学研究所噪声与振动重点实验室,北京100190

出　　处：《网络新媒体技术》2015年第6期7-13,64,共8页Network New Media Technology

摘　　要：论文介绍了一种基于近场噪声监测的远场声压预估方法,讨论了影响该方法预测鲁棒性的主要因素——条件数的影响,并且提出了基于奇异值分解(SVD)的改进计算方法。在此理论基础上,本文开发了一个分布式计算的远场声压预估系统:首先采用Hadoop Streaming工具与python结合的方式,建立了基于MapReduce模型的分布式计算平台;然后在该平台上实现了这种改进的远场声压预估算法,进而实现了分布式计算和科学计算的有机结合;最后搭建了实验所需环境,并在该环境下对预估的结果进行了分析。This paper introduces a far- field sound pressure prediction method based on monitoring near- field noise and discusses the main factor that affects the robustness of the prediction： Condition number. Accordingly an improved method involved with singular value decomposition（ SVD） is proposed. Furthermore a distributed computing sound pressure prediction system has been developed： firstly,a distributed computing platform of MapReduce model is founded with the tool of Hadoop streaming combined with python; then the improved method of far- field sound pressure prediction is realized in this platform which combines distributed computing and scientific computing. Finally,the result of the prediction in our experimental environment has been analyzed.

关 键 词：远场声压预估 HADOOP 分布式计算 STREAMING MAPREDUCE 

分 类 号：TB52[理学—物理] TP311.13[理学—声学]