Semi-Supervised Learning-Based Approach for DOA Estimation Under Hardware Impairments

Abstract

This paper proposes a direction-of-arrival (DoA) estimation algorithm based on semi-supervised learning in the presence of hardware impairments. The proposed algorithm estimates DoA through the following two steps. In the first step, the array response vectors with hardware impairments are estimated by the network version of dictionary learning with un-labeled data. The second step estimates the DoA power spectrum by mapping the DoA and the array response vectors through a small amount of labeled data. Therefore, the proposed algorithm is able to overcome hardware impairments while effectively reducing the labeling cost. Simulation results show that the proposed algorithm maintains high accuracy under severe hardware impairments, which enables practical implementation.