A FTN Signal Detection Method based on BiLSTM Network

• 1、Beijing University of Posts and Telecommunications，Beijing 100876
• 2、School of Information and Communication Engineering,Beijing University of Posts and Telecommunications，Beijing 100876

Abstract：In terms of the complexity of the optimal receiver for FTN associates with the length of channel response and increases exponentially, we present a reduced-classification categories network based on the BiLSTM layer to achieve signal detection. This Network can be considered as a channel equalizer. The BiLSTM detection network adopts the end-to-end mode to recover the original data sources, and the receiver is adaptA FTN Signal Detection Method based on BiLSTM Networkable to different channels and performs noticeably well, without knowing the channel state information in advance. At the same time, according to the characteristics of FTN channel, the unified mapping rule of the real and imaginary parts of the symbol at the sending end is designed, and the duplicate detection symbol is independently detected into the real and imaginary parts in parallel, which not only reduces the classification category but also improves the classification accuracy. In addition, when the inter-symbol interference is too serious, the gain of 1dB can be obtained by appropriately increasing the number of hidden layers at high SNR.

Keywords： FTN Channel Equalization BiLSTM