
Research Article
Spectrum Data Reconstruction via Deep Convolutional Neural Network
@INPROCEEDINGS{10.1007/978-3-030-69072-4_52, author={Xiaojin Ding and Lijie Feng}, title={Spectrum Data Reconstruction via Deep Convolutional Neural Network}, proceedings={Wireless and Satellite Systems. 11th EAI International Conference, WiSATS 2020, Nanjing, China, September 17-18, 2020, Proceedings, Part II}, proceedings_a={WISATS PART 2}, year={2021}, month={2}, keywords={Spectrum data reconstruction Down-sampling Deep convolutional neural network}, doi={10.1007/978-3-030-69072-4_52} }
- Xiaojin Ding
Lijie Feng
Year: 2021
Spectrum Data Reconstruction via Deep Convolutional Neural Network
WISATS PART 2
Springer
DOI: 10.1007/978-3-030-69072-4_52
Abstract
In the paper, we explore the spectrum-data reconstruction of a spectrum-sensing system. In order to decease the demand on the sensed spectrum data, we proposed a deep convolutional neural network (DCNN) based spectrum data reconstruction scheme relying on three stages, thus the satellites are allowed to perform spectrum sensing with the aid of down-sampling, and transmit the low-resolution (LR) and small amount of high-resolution (HR) spectrum data to earth stations. Specifically, in the first stage, the received LR and HR spectrum data will be first preprocessed. Then, the preprocessed HR spectrum data will be sent into the DCNN model for training purposes in the second stage. In the third stage, the preprocessed LR spectrum data will be fed into the trained model with the aid of the optimized hyperparameters, and the trained DCNN can generate the HR spectrum data. Additionally, performance results show that the proposed reconstruction scheme can obtain the reconstructed HR spectrum data in terms of the low mean absolute error.