Deep learning approach for high-resolution DOA estimation

Ch Lokesh Dharma Theja,Kishore Kumar Puli
DOI: https://doi.org/10.1504/ijahuc.2024.138742
2024-05-30
International Journal of Ad Hoc and Ubiquitous Computing
Abstract:The direction-of-arrival (DOA) estimation is a significant problem in array signal processing and their application in different fields ranging from radars to wireless communications. This problem poses significant challenges when the separation of closely spaced sources becomes difficult. The state-of-art methods fail in providing sufficient resolution to separate the DOA's of closely spaced sources unless large number of large number of antennas is being used. In this paper, the problem of resolving direction-of-arrivals (DOAs) of two closely spaced sources using deep learning technology is considered. We propose a novel network named as DAE-CNN-MUSIC network. The problem is framed as a multi-label classification task. The proposed network is trained to make predictions of DOAs across a wide range of signal-to-noise ratios (SNRs), demonstrates enhanced durability in the presence of noise, and resilience even with smaller number of snapshots. The efficiency and superiority of the proposed network is tested under various scenarios. The experimental results demonstrate vast improvement in performance gains at low-SNR regime, and ability to resolve angles with more precision for SNR ≥ 15 dB when compared to state-of-the-art methods.
computer science, information systems,telecommunications
What problem does this paper attempt to address?