Orthogonal projections based algorithm for array signal subspace estimation

Abstract

Subspace-based methods rely on eigenvalue decomposition (EVD) or singular value decomposition (SVD) of a data matrix to compute the array signal or noise subspace, and thus inevitably lead to intensive computational complexity besides a large algorithmic delay. In this paper, an orthogonal projections based algorithm for array signal subspace estimation (OP) is proposed, where by exploiting a reference signal with known waveform the orthogonal projections of array data are calculated step by step to form a set of basis vectors for the signal and/or noise subspace without eigendecomposition. The subspace formed by the OP algorithm is proved mathematically to be equivalent to the real signal subspace. Signal subspace estimation error function and noise subspace estimation error function are defined to evaluate subspace estimation precision. Statistical analysis and simulation results show that the OP algorithm is computationally simple and the subspace estimation precision is roughly equivalent to that of EVD. At the end of the paper, an application to direction-of-arrival (DOA) estimation verifies the good performance of the OP algorithm.