Multi-component filtering of magnetic signals

This paper presents two treatments of real signals recorded by an immersed passive array during an underwater magnetic detection experiment. In this situation, the signal-to-noise ratio is very low, and the noise characteristics are close to the characteristics of the signal that is to be detected. Methods that estimate the signal by minimizing the mean-square distance between the signal and its estimate (using a Wiener filter) are optimal, but some a priori knowledge must first be available before these methods can be used. In this specific application, this a priori knowledge can only be acquired by analyzing the data, and a straightforward application of the optimal method may be misleading. Two filtering methods are therefore implemented to estimate a vector valued signal in a noisy observation. These filtering methods are simple and robust, and they give good results on real data. One method is an adaptation of Wiener filtering. The other consists of global filtering using the eigen-structure of the spectral density matrix of the received signals. Results for each method are presented and compared.