Towards Semi-Supervised Direction Finding With Manifold Regularization
Liuli Wu, Fengyi Tang, Chuan Yu, Xiaoming Liu, Wei Ji, Wenliang Gao · 2024
Machine learning based Direction-of-arrival (DOA) estimation methods heavily depends on the labeled data. When it is difficult to obtain a large number of labeled samples, semi supervised learning can use unlabeled samples to improve the training performance. Therefore, this paper proposes a direction finding method based on semi supervised learning, which uses a small number of labeled samples and a large number of unlabeled samples to gradually modify the DOA estimation function through manifold regularization constraints, so as to improve the DOA estimation performance when labeled data is limited. Simulation results have demonstrated the superiority of the proposed method compared with purely supervised learning when limited labeled samples are available.