用时频峰值滤波方法消减地震勘探资料中随机噪声的初步研究

Reduction of random noise for seismic data by time-frequency peak filtering

时频峰值滤波算法是一种新颖的基于时频分析的信号增强算法,能够有效地消除随机噪声,恢复有效波信息.本文将这种时频分析算法用于消除地震勘探资料中的随机噪声,对淹没于随机噪声下的40道共炮点记录进行时频峰值滤波,恢复出来的共炮点记录可以清楚地表现原始记录同相轴的位置.经过对40道中任选两道(即第21道和第7道)滤波前后的子波形态Wigner-Ville分布、傅立叶振幅谱等的比较,可知仅在谷值和峰值点误差较大,子波带宽相对误差小于25%.仿真试验表明信噪比可达-7dB,说明该方法可以有效地消减地震资料中的随机噪声.

Time-frequency peak filtering （TFPF） is a novel signal enhancement algorithm, which is based on time-frequency analysis. TFPF could eliminate random noise, and recover filtered signal. In this paper, we make use of TFPF to get a clean recovery of common shot records with 40 channels in random noise. It is concluded the errors of peak and valley are bigger and the relative errors of wavelet bandwidth is less than 25M through comparing the wavelet shape, the Wigner-Ville distribution and Fourier transform spectra of amplitude of two filtered channels chosen from the 40 channels at will. Here we choose the twenty-first channel and the seventh channel. The filtered records could clearly show the event of synthetic seismic data, and get a clean recovery of the records in noise level down to a signalto-noise ratio （SNR） of -7dB. Therefore it indicates the efficiency of the algorithm as a noise-eliminated method for seismic data.