Spatio-temporal nonconvex penalty adaptive chirp mode decomposition for signal decomposition of cross-frequency coupled sources in seafloor dynamic engineering

Abstract

Electromagnetic field noise and clutter generated from the motion of ocean waves are the main obstacles in the research of magnetotelluric dynamic analysis, and it is difficult to extract the crossed instantaneous frequencies (IFs) of underwater electromagnetic detected (UEMD) data due to the limited resolution of the current time-frequency techniques. To alleviate this bottleneck issue, a new spatio-temporal nonconvex penalty adaptive chirp mode decomposition (STNP-ACMD) is originally proposed for separating each mono-component individually from a complicated multi-component with severely crossed IFs or overlapped components, in this paper. Specifically, the idea of a nonconvex penalty greedy strategy is incorporated into the vanilla ACMD method by using a recursive mode extraction scheme, and the fractional-order characteristic of the observation signal is also considered. Meanwhile, the spatio-temporal matrices were constructed elaborately and then applied to capture coupling characteristics and spatio-temporal relationships among all estimated mono-components. Eventually, a high-resolution adaptive time-frequency spectrum is obtained according to the IFs and instantaneous amplitudes (IAs) of each estimated mono-component. The effectiveness and practicability of the proposed algorithm were verified via simulated scenarios and velocity dynamic data of the seafloor from the South China Sea, compared with four state-of-the-art benchmarks.