Estimation of source bearings at a randomly deformed array by signal space transformations

A novel approach to array processing is presented which allows high resolution bearing estimation from arrays where the exact sensor locations are not known. It involves rotating weighted eigenvectors in the signal subspace until they resemble wavefront phase delay vectors. In the case of a deformed linear array the phase of these vectors may then be unwrapped and a least squares fit performed. The gradient represents the source wavenumber and the deviations from linear are due to the sensor displacements. The iterative method bears some resemblance to the maximum likelihood solution, but has reduced computational complexity. A simple algorithm involving an eigendecomposition at each iteration is given to present the ideal convergence characteristics. This decomposition is performed on a Hermitian symmetric M×M matrix where M is the number of sources