基于经验模态分解及独立成分分析的微震信号降噪方法

针对微震信号具有高噪声、突变快、随机性强等特点,基于经验模态分解(EMD)及独立成分分析(ICA)提出一种微震信号降噪方法.首先,对含噪信号进行EMD分解,获得一系列按频率从高到低的内蕴模态函数(IMF),利用原信号与各IMF之间的互相关系数辨识出噪声与信号的分界,将分界之上的高频噪声滤除;其次,为有效去除分界IMF中的模态混叠噪声,基于ICA算法对分界IMF进行盲源分离,提取其中的微震有效信号,并将其与剩余的IMF累加重构,从而得到降噪后的微震信号;最后,利用快速傅里叶变换(FFT)时频谱对比分析降噪前后的信号特征,定性说明本文方法的有效性;引入信噪比和降噪后信号占原信号的能量百分比两个参数,定量说明本文方法能充分保留微震信号的瞬态非平稳特征,降噪效果明显.

Micro-seismic signal denoising method based on empirical mode decomposition and independent component analysis

Micro-seismic monitoring system usually works in high noise circumstances, where large amounts of external noise interferes heavily with the successive studies, such as the analysis of micro-seismic arrival time, localization of micro-seismic source, and explanation of earthquake mechanism and so on. Therefore, it's an urgent issue to reconstruct the micro-seismic signal from the polluted signals. Because the micro-seismic signal has characteristics of high noise, fast change and strong randomness, and its bandwidth always overlaps the external noise band in whole or in part, it is difficult to separate the micro-seismic signal from external noise using traditional time-frequency spectrum analysis and classical linear methods. Thus, it is necessary to find a proper de-noising method for micro-seismic signal. Because of the randomness and non-stationarity of micro-seismic signal, a de-noising method is proposed based on Empirical Mode Decomposition (EMD) and Independent Component Analysis (ICA). Firstly, the noisy signal is decomposed by EMD to obtain a series of Intrinsic Mode Function (IMF) ranked by frequency in descending order, and the boundary between noise and signal is identified using the correlation coefficients of the original signal and each IMF, and then the high frequency noises above the boundary are filtered. Secondly, multidimensional inputs are constructed based on the invariability of time translation in order to remove the modal mixing noises in the boundary IMF effectively, and the blind source of the boundary IMF is separated to extract the effective micro-seismic signal. The micro-seismic signal is de-noised through the accumulation and reconstruction of the effective micro-seismic signal and the boundary IMF finally.#br#The following conclusions can be drawn by theoretical and experimental results analysis. (1) The traditional time-frequency spectrum analysis and classical linear methods are poor at de-noising micro-seismic signal which is random and non-stationary. (2) EMD produces modal mix during decomposition due to the strong coupling in time-frequency between micro-seismic signal and the external noise, and Ensemble Empirical Mode Decomposition (EEMD) can inhibit the modal mix to some extent, but its effect is not good enough and it brings the other problems including increase of IMF decomposition and high time complexity. (3) The simulations of noisy Ricker waveform show that SNR of the Ricker is 1.86 dB before de-noising, and it promotes to 16.94dB after using the proposed method, and the energy remains 97.25% of original signal. The effect of this method is obvious. (4) Forty groups of micro-seismic data collected by ISS in May to August 2010 are de-noised by the proposed method, and the results show that the SNR of the noisy signal are promoted from 0 dB to 10~20 dB, with the maximum 19.72 dB (group 25) and the minimum 10.15 dB (group 23). And the energy of de-noised signal remains 89%~99% of the original signal, with the maximum 98.7% (group 8) and the minimum 89.73% (group 13). In order to cope with the modal mix issue in EMD/EEMD method, the micro-seismic are de-noised based on EMD and ICA. EMD is utilized to decompose the noisy micro-seismic signal, and ICA is used to separate the blind source in IMFs with modal mixing noises. The proposed method can remove noise to a greater extent while remaining the more useful information of the origin signal than the other methods.