Research Article
Multi-modem Implementation Method Based on Deep Autoencoder Network
@INPROCEEDINGS{10.1007/978-3-030-69072-4_40, author={Peng Wei and Ruimin Lu and Shilian Wang and Shijun Xie}, title={Multi-modem Implementation Method Based on Deep Autoencoder 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={Satellite communications Deep learning Deep autoencoder network Modem Anti-interception}, doi={10.1007/978-3-030-69072-4_40} }- Peng Wei
Ruimin Lu
Shilian Wang
Shijun Xie
Year: 2021
Multi-modem Implementation Method Based on Deep Autoencoder Network
WISATS PART 2
Springer
DOI: 10.1007/978-3-030-69072-4_40
Abstract
With the fierce competition for electromagnetic spectrum, the development of intelligent satellite communication systems with intelligent waveform generation and reconstruction capabilities is an effective means to adapt satellite communication system to the harsh electromagnetic environment. In this paper, a 10-layer deep autoencoder network (DAN) is designed, and 2-ary to 64-ary modem are implemented based on this 10-layer DAN. During this process, a unified loss function and a unified optimization algorithm are utilized to train and test the 10-layer DAN. Finally, the demodulation performance, that is close to, consistent with or better than that of traditional MPSK or QAM is obtained. The above-mentioned DAN and its training method provide a new way for waveform generation and reconstruction in intelligent communication satellites. In addition, the high-order modulation constellation generated by this 10-layer DAN is quite different from the traditional modulation method and very difficult to distinguish linearly, which is beneficial to improve the anti-intercept ability of the satellite communication waveform.
