Investigation of waveform separation by Independent Component Analysis for composite signals -As a target for mixed signals of earthquake ground motions and railway-induced ground vibrations-

DOI Web Site Web Site 2 Citations 6 References
  • Tsuno Seiji
    Railway Technical Research Institute, Center for Railway Earthquake Engineering Research
  • Iwata Yuichi
    Railway Technical Research Institute, Disaster Prevention Technology Division (Present affiliation is Central Japan Railway Company, Shinkansen Operations Division)

Bibliographic Information

Other Title
  • 独立成分分析を用いた複合信号の波形分離に関する検討<br/>─地震動と列車振動の混合波形を対象として─
  • 独立成分分析を用いた複合信号の波形分離に関する検討 : 地震動と列車振動の混合波形を対象として
  • ドクリツ セイブン ブンセキ オ モチイタ フクゴウ シンゴウ ノ ハケイ ブンリ ニ カンスル ケントウ : ジシンドウ ト レッシャ シンドウ ノ コンゴウ ハケイ オ タイショウ ト シテ

Search this article

Abstract

  Recently, Independent Component Analysis (ICA) has been studied and developed as a new method for multidimensional signal processing in many fields. The ability to detect a certain sound among plural sounds is called the Cocktail-party effect, which is well known as the issue for blind signal separation in acoustics. ICA has been applied to the separation for sounds/biomedical signals and an image processing in the engineering. However, there are not many applications of ICA in fields of exploration geophysics and seismology. To identify noise excited by cars and trains included in seismic data and to separate earthquake ground motions by aftershocks occurred simultaneously during large earthquakes are the issue for blind signal separation; therefore, the development of the Cocktail-party effect is expected in geophysical and seismological fields. As to the separation problem of railway-induced ground vibration mixing in seismic data, in this study, we investigated the waveform separation of earthquake ground motions and railway-induced ground vibrations by ICA, through numerical experiments used the observed data.<br>   At first, we applied ICA to synthetic waveforms mixed earthquake ground motions and railway-induced ground motions spatially without time lag under the condition of BSS, to separate the original signals from those synthetic waveforms. As a result, the earthquake ground motions and the railway-induced ground motions were well separated. And, in the case of distance ratio of 0.99 for two stations from a railway, the earthquake ground motions and the railway-induced ground motions were also well separated. Secondary, we applied ICA to waveforms mixed the observed earthquake ground motions and railway-induced ground motions. As a result, the earthquake ground motions and railway-induced ground motions separated were in good agreement with those observed, especially in the frequency range of 0.5 to 30Hz for the earthquake ground motions. Kurtosis which is the fourth-order cumulant of high order statistics for earthquake ground motions and railway-induced ground motions is large and those frequency distributions are in a shape of super-Gaussian type; therefore, we suggested that ICA is qualified for the separation problem of earthquake ground motions and railway-induced ground motions.<br>

Journal

Citations (2)*help

See more

References(6)*help

See more

Keywords

Details 詳細情報について

Report a problem

Back to top