Minthri Ho

Super-Resolution Techniques in Multistatic Sensor Array Processing

Date: Tuesday, December 6, 2005

Abstract:

Active remote sensing has, for the most part, relied on the monostatic configuration. However, recent developments, such as the DORT method for ultrasonic sensing, serve to highlight the advantages that accrue with multistatic configurations. One gets a much richer set of data by measuring the inter-element frequency response of a multi-static active array, what we call the Frequency Response Matrix (FRM), than by forming the covariance matrix of the signal vector acquired by a passive array. The key property of multistatic antenna arrays that makes them suitable for applying subspace-based signal processing schemes is the rank-deficiency of the FRM This property has been exploited to construct a MUSIC-type target location algorithm, which applies to arbitrary wave-propagation conditions (both mid-field and far-field), and which is much more accurate and robust than previously used time-reversal methods. We analyze the performance of this algorithm in the presence of noise and random perturbations in the target's reflection coefficient. We also derive the statistics of the singular values under the hypotheses of noise only and of target and noise, so that we can evaluate the probability of detection and the probability of false alarm, and obtain ROC curves. We then compare the detection performance of FRM with that of conventional MUSIC. One encouraging conclusion we reach is that these two schemes complement each other on the randomness spectrum of targets. This is because MUSIC relies on the 2nd order moment of the acquired signal, while multistatic FRM uses only the 1st order moment (or even just a single snapshot) to estimate target location. We also show that the subspace-based FRM approach, which has been developed explicitly for a small number of point scatterers, can be applied also to extended targets. We present both a direct (non-parametric) method and parametric method based on truncated Taylor-series approximation.

Committee:
Prof. Vinay Ingle
Prof. Hanoch Lev-Ari (advisor)
Prof. Eric Miller