Fast Matched Filter (FMF): An Efficient Seismic Matched‐Filter Search for Both CPU and GPU Architectures

• DOI: 10.1785/0220170181
• Publication Date: 2017-12-20T22:41:53Z
• Authors: Eric Beaucé; William B. Frank; Alexey Romanenko
• Keywords: 0103 physical sciences; 01 natural sciences

Research Article| December 20, 2017 Fast Matched Filter (FMF): An Efficient Seismic Matched‐Filter Search for Both CPU and GPU Architectures Eric Beaucé; Beaucé aDepartment of Earth, Atmospheric, Planetary Sciences, Massachusetts Institute Technology, 77 Avenue, 54‐527, Cambridge, 02139 U.S.A., [e-mail] other works by this author on: GSW Google Scholar William B. Frank; Frank Alexey Romanenko bChair Informatics Systems, Novosibirsk State University, Pirogova, 2, 630090, Russia Author Article Information Publisher: Seismological Society America First Online: 20 Dec Online Issn: 1938-2057 Print 0895-0695 © Letters (2018) 89 (1): 165–172. https://doi.org/10.1785/0220170181 history Cite View This Citation Add to Manager Share Icon Facebook Twitter LinkedIn MailTo Tools Get Permissions Site Beaucé, Frank, Romanenko; Architectures. 2017;; doi: Download citation file: Ris (Zotero) Refmanager EasyBib Bookends Mendeley Papers EndNote RefWorks BibTex toolbar search Dropdown Menu input auto suggest filter your All ContentBy SocietySeismological Advanced ABSTRACT Matched‐filter searches are an important tool in modern seismology detect seismic events. They operate via algorithm that computes the correlation coefficient between a template event sliding window continuous records. A detection is recorded when crosses established threshold. We present optimized program, called (FMF), efficiently runs network‐based matched‐filter with either central processing units (CPUs) or Nvidia graphics (GPUs). Wrappers both Python MATLAB (CPU only) provided easily run FMF on wide range computational resources, from multicore laptops specialized computing clusters GPUs. implementations leverage significantly similar structure it comes computation coefficients time domain achieve rapid performance. The highly parallel architecture GPUs lends itself perfectly algorithm, we fastest times our implementation. allows matched‐filtering large set waveforms dataset reasonable amount time. Such fast step expanding scope earthquake fostering reproducibility such studies. You do not have access content, please speak institutional administrator if you feel should access.