Matching Pursuit with Enhanced Performance and Adaptability

One of the advantages which matching pursuit (MP) method has over conventional time-frequency decomposition is its ability to get capture of time-frequency fingerprint with extremely high resolution, which also makes itself quite sensitive to data noises and results in the non-uniqueness of its solution. In order to further diminish the influence of noises and stabilize the performance, two targeted modifications are incorporated in this work. Specifically, the lateral continuity of the underground layers is exploited as a constraint and as a reference for aligning the signal traces in the decomposition process. As a consequence, both the number of decomposed wavelets and the computation time can be significantly reduced. Moreover, choosing the best wavelet in each iteration follows a minimal residual energy criterion, which further results in a smaller number of resultant wavelets and thus makes the decomposition more robust.