Low-complexity DOA Estimation via Subspace Rotation Technique

The MUltiple SIgnal Classification (MUSIC) algorithm is one of the most important techniques for Direction-Of-Arrival (DOA) estimate. However, this method is found expensive in practical applications, due to the heavy computational cost involved. To reduce the complexity, a novel efficient estimator based on Subspace Rotation Technique (STR) is proposed. The key idea is to divide the noise subspace matrix along its row direction into two sub-matrices, and perform STR to get a new rotated sub-noise subspace with reduced dimensions. As this rotated sub-noise subspace is also orthogonal to the signal subspace, a new cost function is finally derived to estimate DOAs. Theoretical analysis indicates that redundancy computations in spectral search are efficiently avoided by the proposed method as compared to MUSIC, especially in scenarios where large numbers of sensors are applied to locate small numbers of signals. Simulation results verify the effectiveness and efficiency of the new technique.