Direction of arrival estimation in passive radar based on deep neural network
Abstract Most traditional direction of arrival (DOA) estimation methods in passive radar are based on the parametric model of the antenna array manifold, and lack the adaption to the array errors. The data‐driven machine learning‐based methods have great array error adaption capability. However, mos...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2021-12-01
|
| Series: | IET Signal Processing |
| Subjects: | |
| Online Access: | https://doi.org/10.1049/sil2.12065 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850225789802381312 |
|---|---|
| author | Xiaoyong Lyu Jun Wang |
| author_facet | Xiaoyong Lyu Jun Wang |
| author_sort | Xiaoyong Lyu |
| collection | DOAJ |
| description | Abstract Most traditional direction of arrival (DOA) estimation methods in passive radar are based on the parametric model of the antenna array manifold, and lack the adaption to the array errors. The data‐driven machine learning‐based methods have great array error adaption capability. However, most existing machine learning‐based methods cannot be applied directly to the passive radar DOA estimation, because the array covariance matrix that they use as the input is not easy to estimate with adequate accuracy in passive radar owing to the poor target signal to clutter plus noise ratio (SCNR). A deep learning‐based method for DOA estimation in passive radar is proposed here. Clutter cancelation and range–Doppler cross‐correlation (RDCC) is performed to increase the target SCNR. The RDCC result is taken as the input of the deep learning method, and the amplitude and phase uncertainties of the RDCC result are treated. A two‐stage deep neural network (DNN) is designed. The first stage determines the spatial sub‐region of the target, and the second stage gets the refined DOA estimation. Simulations show that the proposed two‐stage DNN well outperforms the traditional passive radar DOA estimation method and the multi‐layer perceptron network. Real experiments verify the superiority of the proposed method. |
| format | Article |
| id | doaj-art-5bceb2daf5164291b9f8a8841f4b7316 |
| institution | OA Journals |
| issn | 1751-9675 1751-9683 |
| language | English |
| publishDate | 2021-12-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Signal Processing |
| spelling | doaj-art-5bceb2daf5164291b9f8a8841f4b73162025-08-20T02:05:14ZengWileyIET Signal Processing1751-96751751-96832021-12-0115961262110.1049/sil2.12065Direction of arrival estimation in passive radar based on deep neural networkXiaoyong Lyu0Jun Wang1School of Information Engineering Zhengzhou University Zhengzhou ChinaNational Laboratory of Radar Signal Processing Xidian University Xi’an ChinaAbstract Most traditional direction of arrival (DOA) estimation methods in passive radar are based on the parametric model of the antenna array manifold, and lack the adaption to the array errors. The data‐driven machine learning‐based methods have great array error adaption capability. However, most existing machine learning‐based methods cannot be applied directly to the passive radar DOA estimation, because the array covariance matrix that they use as the input is not easy to estimate with adequate accuracy in passive radar owing to the poor target signal to clutter plus noise ratio (SCNR). A deep learning‐based method for DOA estimation in passive radar is proposed here. Clutter cancelation and range–Doppler cross‐correlation (RDCC) is performed to increase the target SCNR. The RDCC result is taken as the input of the deep learning method, and the amplitude and phase uncertainties of the RDCC result are treated. A two‐stage deep neural network (DNN) is designed. The first stage determines the spatial sub‐region of the target, and the second stage gets the refined DOA estimation. Simulations show that the proposed two‐stage DNN well outperforms the traditional passive radar DOA estimation method and the multi‐layer perceptron network. Real experiments verify the superiority of the proposed method.https://doi.org/10.1049/sil2.12065antenna arraysarray signal processingcovariance matricesdirection‐of‐arrival estimationlearning (artificial intelligence)multilayer perceptrons |
| spellingShingle | Xiaoyong Lyu Jun Wang Direction of arrival estimation in passive radar based on deep neural network IET Signal Processing antenna arrays array signal processing covariance matrices direction‐of‐arrival estimation learning (artificial intelligence) multilayer perceptrons |
| title | Direction of arrival estimation in passive radar based on deep neural network |
| title_full | Direction of arrival estimation in passive radar based on deep neural network |
| title_fullStr | Direction of arrival estimation in passive radar based on deep neural network |
| title_full_unstemmed | Direction of arrival estimation in passive radar based on deep neural network |
| title_short | Direction of arrival estimation in passive radar based on deep neural network |
| title_sort | direction of arrival estimation in passive radar based on deep neural network |
| topic | antenna arrays array signal processing covariance matrices direction‐of‐arrival estimation learning (artificial intelligence) multilayer perceptrons |
| url | https://doi.org/10.1049/sil2.12065 |
| work_keys_str_mv | AT xiaoyonglyu directionofarrivalestimationinpassiveradarbasedondeepneuralnetwork AT junwang directionofarrivalestimationinpassiveradarbasedondeepneuralnetwork |