对依据频率的匹配追踪快速算法的改进

An improvement for frequency-dominated fast matching pursuits algorithm

(1.中国中煤能源集团公司信息技术管理中心，北京100120； 2.中国科学院地质与地球物理所,北京100029)

Information and Technology Management Center, China National Coal Group Corporation, Beijing 100120, China

匹配追踪（MP）算法是信号自适应分解的一种具体实现方法。依据频率的匹配追踪快速算法利用地震信号有限频带的先验知识，建立以频率为主要搜索目标的字典索引，减少了字典索引和迭代循环的计算量，提高了信号匹配追踪分解的效率和计算速度。改进算法以复地震信号能量极值处的瞬时频率为初始频率，建立频率与尺度参数的关联关系，使循环迭代过程紧密围绕残值信号能量的有效衰减，提高了确定初始频率的针对性。

Matching pursuits (MP) algorithm is a specific implementing method for signal self-adapted decomposition. With the prior-knowledge that the seismic signal is composed of certain limited frequency band, the frequency-dominated fast MP algorithm creates dictionary index which regards frequency as main matching target, so that it can decrease the amount of dictionary index size and iteration computation, and promote the efficiency and computation speed of signal MP decomposition. By taking the instantaneous frequency at the supreme points of complex seismic signal as original frequency and establishing the relation of linking frequency with scale parameters, the MP iteration process tightly relates with the effective attenuation of residual signal, which improves the pertinence for identifying original frequency.