CAJ | 학술논문

Welch算法是一种应用很广的经典功率谱估计算法。但是面对现在日益膨胀的海量数据，单纯在串行Mat-lab环境下运行Welch算法势必耗费大量的运算时间。尽管Matlab也引入了并行计算工具箱，但是价格昂贵，不利于大范围推广使用。根据Welch算法的原理，在Linux集群环境及消息传递接口MPI 的支持下，采用主从并行编程模式，实现了一个开源的Welch并行算法PMWelch。实验结果表明，PMWelch不仅具有Matlab下Welch算法一样的运算结果，还可以大幅减少运行时间。
Welch산법시일충응용흔엄적경전공솔보고계산법。단시면대현재일익팽창적해량수거，단순재천행Mat-lab배경하운행Welch산법세필모비대량적운산시간。진관Matlab야인입료병행계산공구상，단시개격앙귀，불리우대범위추엄사용。근거Welch산법적원리，재Linux집군배경급소식전체접구MPI 적지지하，채용주종병행편정모식，실현료일개개원적Welch병행산법PMWelch。실험결과표명，PMWelch불부구유Matlab하Welch산법일양적운산결과，환가이대폭감소운행시간。
It is well known that Welch algorithm is one of classical power spectrum estimation which is used widely. But facing a rapidly growing massive data, it will spend a plenty of time to operate Welch algorithm in the serial Matlab circum-stances. Though a parallel computing toolbox is involved in Matlab in 2004, it is too expensive to be used widely. A parallel Welch algorithm, which is implemented using the Master-Slave parallel programming model with the support of Linux clusters and MPI, is proposed here depending on the theory of Welch algorithm. The experiment results show that PMWelch can achieve the same accuracy as Matlab do and can highly improve the speed.