期刊
AEU-INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATIONS
卷 173, 期 -, 页码 -出版社
ELSEVIER GMBH
DOI: 10.1016/j.aeue.2023.155008
关键词
Automatic modulation classification; DOA estimation; Convolutional neural network; Intelligent radio communication
This paper proposes a method that combines UNet and rootMUSIC algorithm to improve the accuracy of DOA estimation. By converting the covariance matrix of received signals into an ideal antenna array's covariance matrix, and then using rootMUSIC algorithm for DOA estimation, the experimental simulation results show that this method can significantly reduce the error of DOA estimation compared with traditional methods.
In the practical direction-finding systems, the accuracy and resolution of the direction of arrival (DOA) estimation are affected not only by the Gaussian noise and array size but also by hardware configuration imperfections, such as errors in element manufacturing and mounting. These impairments cause phase and amplitude errors in estimating the DOA of signal sources. To address this issue, this paper proposes to combine a U-shape deep neural network (UNet) with the multiple signal classification via the root of the polynomial (rootMUSIC) algorithm (so-called UNet-rootMUSIC) to improve the DOA estimation accuracy. In this approach, the UNet model plays a role in converting a covariance matrix of received signals containing phase and gain errors into a nearly perfect one of the ideal antenna array. The rootMUSIC algorithm is then employed to estimate the signal DOA based on the converted covariance matrix. The DOA estimation performance of the uniform linear array of eight elements with an inter-element distance of lambda/2 is analyzed through experimental simulations. The simulation results demonstrate that our method can significantly reduce the root mean square error of DOA estimation compared to the conventional MUSIC, rootMUSIC, ESPRIT methods and two deep neural network-based angular classification methods.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据