[C08] BiLSTM-Based Frame Synchronization for Overlapped S-AIS Signals: A Learning-Empowered Approach

In this paper, focusing on the signal detection of space-borne automatic identification system (S-AIS), two learning-empowered frame synchronization methods are proposed, which predict the accurate overlapping position of two S-AIS signals with the help of a bidirectional long short-term memory (BiLSTM) network. In particular, by regarding the frame synchronization as a binary classification issue, BiLSTM network can be utilized to find the overlapping position of the received signals accurately. Furthermore, convolutional neural network (CNN) is introduced into the proposed BiLSTM-based approach to handle the non-smooth power fluctuation. Simulation results show that our proposed learning-empowered methods outperform the conventional frame synchronization method in terms of accuracy and robustness, which can work effectively even under various communication conditions.

Recommended citation: T. Yang, D. He, Z. Lu, H. Wang, H. Zhao and Z. Wu, "BiLSTM-Based Frame Synchronization for Overlapped S-AIS Signals: A Learning-Empowered Approach," in Proc. 2023 IEEE/CIC International Conference on Communications in China (ICCC), Dalian, China, 2023, pp. 1-6.
Download Paper