| 摘要: |
| 针对无源定位中噪声统计特性不准确和对多源信息的综合利用，提出一种利用深度神经网络（DNN）的无源定位方法，该方法将训练集数据输入到深度神经网络中进行学习训练，利用随机失活这一正则化方法提高了模型的泛化能力，对模型的超参数选择进行二维搜索，最终得到深度神经网络模型的最优参数设置。将其和传统的无源定位方程解算方法以及单层神经网络模型进行对比，仿真结果表明提出的方法能有效降低噪声对无源定位的精度影响，增强了系统鲁棒性，同时也证明了深度神经网络对多源信息的综合利用能力。 |
| 关键词: 无源定位 多源信息 深度神经网络（DNN） 鲁棒性 |
| DOI：10.3969/j.issn.1672-2337.2018.04.012 |
| 分类号:TN958.97; TP183 |
| 基金项目: |
|
| Research on Application of Deep Neural Network in Passive Location |
|
LIU Yu， FENG Sheng, WANG Guiling
|
|
Beijing Institute of Radio Measurement, Beijing 100854, China
|
| Abstract: |
| Aiming at the inaccuracy of noise statistics in passive location and the comprehensive utilization of multi-source information, a passive localization method using deep neural network is proposed. The training set data are input into the deep neural network (DNN) for training. The regularization method of dropout improves the generalization ability of the model. The optimal parameter setting of the DNN model is finally obtained through searching for the hyperparameter of the model in two dimensions. The simulation results show that the proposed method can effectively reduce the impact of noise on the accuracy of passive localization and enhance the robustness of the system compared with the traditional method of solving passive localization equation and single-layer neural network model. Meanwhile, it proves that the DNN model can comprehensively utilize multi-source information. |
| Key words: passive location multi-source information deep neural network(DNN) robustness |
