Joint Self-Interference Cancellation and Data Estimation for OFDM Based Full-Duplex Communication Systems
Sprache des Vortragstitels:
Englisch
Original Tagungtitel:
Asilomar Conference on Signals, Systems and Computers (ACSSC 2023)
Sprache des Tagungstitel:
Englisch
Original Kurzfassung:
In this work, we study joint self-interference cancellation (SIC) and data estimation in a full-duplex (FD) communication system utilizing neural networks (NNs). The proposed approaches are developed for an FD communication system with 5G conform OFDM transmit and receive signals and a data transmission over a multipath channel. We investigate three different options for conducting SIC and data estimation with NNs. First, we aim to train two NNs consecutively, the first one performing SIC in time domain and the second NN performing data estimation in frequency domain. Here, we follow a two-step approach where the NN conducting SIC is trained initially, and after freezing its trained parameters the NN for data estimation is optimized. In the second setup, both NNs are trained jointly in an end-to-end manner, while the basic structure of the first setup remains the same. In contrast to that, for the third scenario, we employ only one NN for both SIC and data estimation. We compare the aforementioned approaches in terms of achievable BER performance and computational complexity with a state-of-the-art model-based approach, where a polynomial SI canceller and a zero-forcing OFDM data estimator are applied.
Forschungs-