A sparse-based DOA estimation method based on deep learning
Tao Luo, Peng Chen, Zhimin Chen, Zhenxin Cao · IET conference proceedings. · 2024
With the rapid advancement of assisted driving and automated driving, there is an increasing urgency for a super-resolution direction of arrival (DOA) estimation method in vehicular sensing systems. In this paper, a high-accuracy DOA estimation algorithm based on deep learning is proposed. The real and imaginary parts of normalized single snapshot received signals are used as the input of the proposed network, which makes the network robust. Different from the other existing DOA estimated network, the output of proposed is a vector, and the spatial spectrum is obtained after a simple transformation, leading to a smoother spatial spectrum and improved accuracy. The simulations are performed to demonstrate its effectiveness, and a frequency-modular continuous wave (FMCW) radar location system is established to validate it in practice.