Space-time MUSIC imaging of EMI sensing data and a subspace partition study

To localize sources in electromagnetic induction (EMI) sensing, we propose a space-time MUltiple SIgnal Classification (MUSIC) imaging approach. Under the EMI physical model, the approach is established upon a receiver array-based spatial and temporal response matrix. In contrast to the multi-static response matrix-based MUSIC imaging approach that requires a sufficient number of receivers and transmitters, the new imaging scheme relaxes the condition on a large number of transmitters and thus becomes applicable to a wide range of EMI sensing systems. We also investigate the issue of forming a noise subspace that is essential to the MUSIC performance. Our theoretic analysis indicates that the choice of an underestimated noise subspace is appropriate for the imaging. Thus it implies, when doing a subspace partition required in the MUSIC, that a difficult task of exactly distinguishing signal from noise eigenvalues can be eased greatly in a practical eigenvalue distribution map. The technique is evaluated through the synthetic and real data. The results show that the space-time MUSIC can be used to detect sources with an EMI system consisting of multiple receivers but a few transmitters. As predicted from the theoretical analysis, the tests show that the MUSIC-source locations are robust to under-estimates of the dimension of the noise subspace.