Ricker子波核最小二乘支持向量机在地震勘探信号去噪应用中的参数设置研究

Study of parameters setting for least square support vector machine based on Ricker wavelet kernel in the denoising applications of seismic prospecting signals

最小二乘支持向量机(LS-SVM)用于拟合回归处理时的参数设置一直是一个难题,它会受到信号类型和强度、核函数类型、噪声强度、计算精度要求等因素的影响.本文针对Ricker子波核LS-SVM去除地震勘探信号中随机噪声问题,讨论和分析了向量机参数、核参数对去噪性能的影响.实验表明,核参数f可取为地震记录的主频,不能较准确估计时宁大勿小;向量机参数γ只要不取得过小,一般情况下都是能接受的.采用此方法对含不同强度噪声的地震勘探信号进行了去噪处理.

Parameters setting for LS-SVM used in the regression processing is a difficult problem at all times. This setting is affected on the type and amplitude of signals, the type of kernel function, intensity of noise and calculation precision, etc. The setting of SVM parameter and kernel parameter in the denoising applications of seismic prospecting signals is analyzed and discussed separately in this paper. The experimental results show that kernel parameter f can be selected as the predominate frequency of the seismic prospecting records, bigger instead of smaller if not estimated accurately, and the acceptable range of the selection of SVM parameter γ is very wide except too small. According to the above methods of parameters setting, the simulation experiments on the noisy seismic prospecting signals have been done .