Skip to content

利用稀疏去噪自编码器和深度神经网络实现小型无人机的波达方向估计

License

Notifications You must be signed in to change notification settings

DNNlearner/DeepDOA

 
 

Repository files navigation

DeepDOA

基于深度神经网络的稀疏降噪自编码器SDAE在小型无人机的测向实现。与子空间分解类算法ESPRIT和MUSIC算法进行比较。

由于相位同步机制、天线校准机制和天线辐射模式的分析模型都不是必要的,所以所提出的DF方案是实用且低复杂度的。此外,所提出的DF方法频偏分析方法可使用单通道射频接收机来实现。

For more details, please see our Arxiv paper.

Whole Architecture:

Architecture training phase:

Dependencies

  • Tensorflow (recommended below 1.5)
  • Numpy 1.14.4

Dataset数据集

提供部分数据演示所提出的方法 训练数据: Dround_Data_New/Normalized 测试数据 : Dround_Data_New/Normalized_test

数据以45度进行区域划分,总计有8个区域。 例如:deg_0_normalize.csv数据文件代表从第一个区域收集到的训练文件。

File description 文件描述

  • DNN_Ground_data_8sectors.py : Implementation without SDAE
  • DenoisingAE.py :实现稀疏降噪自编码进行单独训练,学习降噪特征 Implementation of SDAE for training it separately to learn denoising features.
  • get_csv_data.py : Data handler
  • main.py : 结合稀疏降噪自编码SDAE和DNN实现DOA估计 combining SDAE with a neural network to perform DOA estimations

Citation引证

If this is useful for your work, please cite our Arxiv paper:

@article{abeywickrama2017rf,
  title={RF-Based Direction Finding of UAVs Using DNN},
  author={Abeywickrama, Samith and Jayasinghe, Lahiru and Fu, Hua and Yuen, Chau},
  journal={arXiv preprint arXiv:1712.01154},
  year={2017}
}

License

This is released under the MIT license. For more details, please refer LICENSE.

"Copyright (c) 2018 Lahiru Jayasinghe"

About

利用稀疏去噪自编码器和深度神经网络实现小型无人机的波达方向估计

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 99.9%
  • Hack 0.1%