计算机科学
到达方向
天线阵
编码器
无线
雷达
变压器
算法
天线(收音机)
职位(财务)
高分辨率
实时计算
传感器阵列
电子工程
信号处理
维数(图论)
多信号分类
信号(编程语言)
多径传播
无线传感器网络
作者
Tiezhen Jiang,Q. H. Li,Runjie Kang,Qingzhu Li
标识
DOI:10.1088/1361-6501/ae264c
摘要
Abstract Direction of arrival (DOA) estimation plays a vital role in automotive radar, speech recognition, and wireless communication applications. Most existing DOA estimation algorithms rely on uniform linear arrays (ULA) with equally spaced elements and simple structures. However, in real-world scenarios, random sensor failures occur, causing the ULA to degrade into a randomly sparse linear array (SLA). The missing data from faulty sensors significantly reduces the performance of traditional DOA estimation methods. To address this issue in radar systems, this paper proposes a novel DOA estimation approach called TAPEDOA-Net, which is based on a Transformer Adaptive Positional Encoder (TAPE). This method preserves the spatial array dimension features of the received signals during feature extraction. It scales the standard Transformer positional encoding using array position information. The Transformer encoder effectively integrates signal features with array position information. This integration reduces the impact of antenna faults on DOA estimation performance. The simulation results show that, compared to several existing methods, the proposed TAPEDOA-Net achieves accurate DOA estimation for both intact ULA and ULA with known failure positions. It achieves higher angular resolution and lower angular error in these complex environments.
科研通智能强力驱动
Strongly Powered by AbleSci AI