DoA Estimation for Lens Antenna Array via Root-MUSIC, Outlier Detection, and Clustering

Lens antenna arrays have become attractive for mmWave MIMO systems due to their energy-focusing and path-division properties. However, when the signals cannot be well differentiated at different array elements due to their similar incident angles, we cannot estimate their DoAs separately. In this paper, we present a DoA estimation algorithm for this situation. Our algorithm has three main steps: a special version of root multiple signal classification (MUSIC) for lens antenna arrays, outlier detection, and clustering. Numerical results show that our algorithm can achieve good performance even with a large number of signal sources, a large number of array elements, and a small number of snapshots.