Self-adaption independent component analysis extraction method of weak signals of microseism

Abstract

The invention discloses a self-adaption independent component analysis extraction method of weak signals of microseism. The method comprises the steps that each microseism record is preprocessed, and then eigenvalue analysis is conducted so that the number of independent components can be determined; separation is conducted on observation signals according to the ICA algorithm based on kurtosis, so that an initial estimated value of a source signal is obtained; filtering is conducted on the estimated value of each independent component through an improved self-adaption prediction method, and then a self-adaption prediction filtering result of each independent component is obtained; in the process of calculation, fitting is conducted on one loop iteration error and a weight coefficient through a curve fitting method, and a fitting result is substituted into a primary formula for calculation of a next cycle; finally, weight fusion is conducted on the self-adaption prediction filtering results of all the independent components, and the optimal prediction result is obtained. By the adoption of the self-adaption independent component analysis extraction method of the weak signals of the microseism, acute oscillation of errors is reduced, an algorithm result can be stable as expected, the weak signals in the microseism records can be effectively extracted, and the signal to noise ratio of the weak signals is improved.

Description

technical field [0001] The invention relates to a microseismic detection technology, in particular to a microseismic self-adaptive independent component analysis weak signal extraction method. Background technique [0002] The temporal and spatial variability of microseismic signals determines the variability of signal processing methods. Although the filtering based on Wiener theory does not require the input of prior information, the filter is fixed during the filtering process; the filtering method of Kalman filtering theory requires the prior information of the expected output, but the actual microseismic signal due to the diversity of changes , it is difficult to statistically verify a fixed expected output, so it is limited in practical application, and the effect of data processing is not satisfactory. Adaptive filtering automatically adjusts and updates the expected output, and finally obtains the best expected output estimate through multiple iterations, overcoming...

AI Technical Summary

Problems solved by technology

For microseismic signals, the monitoring problem of such weak signals similar to random pulses in the time direction becomes more difficult

Existing methods for weak signal extraction using signal periodicity characteristics are limited

Images