基于多核模型的地震信号高效稀疏分解

为了提高地震信号分解算法的效率和模型的稀疏度, 本文利用多个核函数作为原子, 自适应地对地震信号进行稀疏分解。通过对地震信号在时频域分别进行全局k均值聚类, 确定字典库中原子备选参数, 然后通过正交最小二乘算法进行信号的稀疏重构。合成资料以及实际地震资料应用结果均表明, 文中所提方法在达到同样的重构精度时, 较大程度地提高了地震信号分解的稀疏度。

Efficient seismic sparse decomposition based on multiple kernel-based models

To enhance the efficiency and sparsity of seismic signal decomposition, multiple kernels are used for the adaptive sparse decomposition of seismic signals. At first, the global k-means clustering algorithm is utilized to generate the preselected behavioral parameters in the dictionary. Then the signal is reconstructed with orthogonal least squares method. The experiments both on synthetic and real data were conducted to evaluate the performance. The results show that multiple kernel-based models greatly improve the sparsity with the similar accuracy.