A Combined Joint Diagonalization-Music Algorithm For Estimating Locations Of Subsurface Targets

In this paper we present a method for extracting target location from data measured by Time- Domain Electromagnetic Multisensor Towed Array Detection System (TEMTADS). The TEMTADS is consisted of square-loop transmitters and 3-D receivers, arranged in a 2×2 array, which generates multistatic response (MSR) matrices of 4×12 in Ng time channels. The data collected first goes through a linear combination process to form a square matrix so we can apply joint diagonalization (JD), a technique that finds the eigenvectors which diagonalize the entire set of MSR matrices. A filtering process is embedded in the JD to enhance signal-to-noise ratio (SNR). The eigenvectors arouse from Targets of interest (TOI) and from noise can be identified after applying JD, and this information is passed to a multiple signal classification (MUSIC) algorithm to separate the signal and noise subspaces. MUSIC algorithm then projects the noise subspace onto a theoretically calculated Green’s Function array. Due to their orthogonality, the target locations can be estimated by looking for maximums when we invert this multiplication. With the Green’s Function pre-calculated, the method can be carried out fast enough to perform targets mapping in real or near real-time.