Neural network-based time-domain equalization without training signal in OFDM systems without CP
Kai Isaka, Teruyuki Miyajima, Yoshiki Sugitani · IEICE Communications Express · 2021
This paper proposes a neural network-based time-domain equalizer (TEQ) in OFDM systems. The proposed TEQ based on the minimum output energy criterion does not require the transmission of a training signal or the insertion of a cyclic prefix to suppress inter-symbol interference; thus, the proposed TEQ does not degrade bandwidth efficiency. Further, an arbitrary decision delay and multiple receive antennas are introduced to improve the bit error rate performance. By simulation, we show that the proposed TEQ is significantly superior to a conventional scheme.