Random Noise Suppression Algorithm for Seismic Signals Based on Principal Component Analysis

Seismic data may suffer to serious noise signal, therefore it's necessary to further process and interpret it. In this passage, we proposed a new method about noise suppression for seismic data based on principal component analysis (PCA), including following four steps. Firstly, one-dimensional seismic signals are extended to multidimensional dataset. Secondly, to de-correlate the new dataset, we use Gaussian noises to whiten the generalized signals with the signal noise ratio (SNR) of noises equalling to the data SNR. Thirdly, with regard to the uncorrelated dataset, we execute random noise suppression using PCA technology from transform domains, which is spanned by the eigen-vector of the data co-variance matrix. Finally, interesting data of seismic data is changed back to time domain by corresponding inverse transform. We confirmed the effectiveness of the proposed method by simulation results of measurements data and seismic signals.