Efficient computation of the eigensubspace for array signal processing

The authors present an efficient technique for computing the eigensubspace spanned by a received array data vector. First, the sensor array is partitioned into several subarrays without overlapped sensors. The basis matrix for the signal subspace of each subarray is computed. Using these basis matrices, an additional subarray is constructed and the basis matrix for the corresponding signal subspace is also computed. By using these subarray basis matrices, the basis matrix for the signal subspace of the original sensor array can be computed with reduced computing cost. The statistical performance and computational complexity for adaptive beamforming and for bearing estimation using the proposed technique are evaluated. The theoretical results are confirmed and illustrated by simulation results