Multivariate Fast Iterative Filtering and Intrinsic Mode Functions for Time Delay Estimation Applied to Motion Estimation for Synthetic Aperture Sonar Imagery

Abstract

Synthetic Aperture Sonar (SAS) provides the best opportunity for side-looking sonar to achieve high-resolution images of underwater objects. However, SAS processing requires maintaining a coherent phase history over the entire synthetic aperture, driving strict constraints on resolvable platform motion. This has compelled the development of motion estimation and compensation techniques that use the received ping data, in addition to the onboard navigation solution, to resolve ping-toping platform motion. The most common approach is to use the redundant phase center (RPC) technique whose accuracy depends upon a Time Delay Estimation (TDE) calculation between two redundant pings. Given the stochastic nature of the undersea operational environment, some level of decorrelation between these two signals is likely. This could result in an inaccurate time delay estimation. In this research two iterative signal decomposition methods, well suited for non-linear and non-stationary signals, are investigated for their potential to improve the TDE; the Multivariate Fast Iterative Filtering (MFIF) Algorithm and the Fast and Adaptive Multivariate Empirical Mode Decomposition (FA-MVEMD) algorithm. The TDE performances for simulated sonar signals, over a wide range of SNR, are presented. At a very low SNR, −15dB, knowing the theoretical value of the time delay, the results show a decrease in TDE error of for both algorithms used to decompose the signals in preprocessing, as compared with the Baseline estimation error, in which the time delay is estimated using the original non-decomposed signals. There was a 47 percent decrease for the MFIF and 15 percent decrease for the FA-MVEMD.