Independent Component Analysis And Its Application On Blind Separation Of Seismic Data

In seismic exploration,seismic data usually contains random noise and it reduces the SNR(Signal Noise Ratio),which directly reduces the reliability of seismic data and the accuracy of parameter extraction. Improving the SNR is almost throughout each aspect in seismic data processing. So how to improve the SNR of seismic data gradually becomes a goal. Independent component analysis (ICA) technology is a new technology method of blind source separation which develops in recent years. ICA is a new multi-dimensional signal processing method based on level statistics,also is able to achieve separation of source signals in the absence of priori information. Observed signal will be used to establish the objective function in accordance with the principles of statistics independence. Through the optimization algorithm, observed signal will be divided into a number of independent signal components in order to extract the valuable signal.The seismic signals and random noise generally agree with the precondition of ICA. So this paper studies the method of the elimination of random noise in the seismic exploration applying BSS model of ICA.Then we present the ICA algorithm based on maximum negentropy and improve it,including iteration pattern and pretreatment. This paper presents the improved de-noising algorithm using the features of adjacent seismic data each other. For a group of multichannel seismic data,concretely operating steps can be summed as follows.(1) The two adjacent channel of seismic data will be inputted the separation system starting from the first channel. Extracting the pure sources by adopting the double overlap processing adjacent seismic data, which can avoid the losing of the valuable data. In other words,every adjacent overlap data are inputted the separation system.(2)We do the pretreatment including mean removal and whiting, Compared to the common PCA (principal component analysis),the paper try to alter whiting based on SVD(singular value decomposition).Not only maximumly remove the additive noise, but also meet the demand of ICA for the primary signal.(3)For the seismic data have been got by the previous step,the paper uses the ICA model base on negentroy maximum for blind separation of two results ,random noise and the valuable data.(4) Because of ICA nondeterminacy,we must sort out the separated data. The paper sets up a matching coefficient to identify effective signal to resolve the order uncertainty problem in ICA, then achieve an effective signal extraction. Moreover, the model is more robust and more accurate than traditional algorithm.Finally,simulation experiments using simulate signal make performance comparison and analysis. The experimental results demonstrate that the algorithm is feasible. Meanwhile, separating the practical seismic data show that the model proposed in this paper effectively remove the random noise.