Signal detection using multi-channel seismic data

Abstract

Abstract We outline a simple signal detection approach for multi-channel seismic data. Our approach is based on the premise that the wave-field spatial coherence increases when a signal is present. A measure of spatial coherence is provided by the largest eigenvalue of the multi-channel data's sample covariance matrix. The primary advantages of this approach are its speed and simplicity. For three-component data, this approach provides a more robust statistic than particle motion polarization. For array data, this approach provides beamforming-like signal detection results without the need to form beams. This approach allows several options for the use of three-component array data. Detection statistics for three-component, vertical-component array, and three different three-component array approaches are compared to conventional and minimum-variance vertical-component beamforming. Problems inherent in principal-component analysis (PCA) in general and PCA of high-frequency seismic data in particular are also discussed. Multi-channel beamforming and the differences between principal component and factor analysis are discussed in the appendix.