LSTM-Autoencoder For Faster-Than-Nyquist Optical Wireless MIMO Communication System

Faster-than-Nyquist (FTN) technology introduces additional inter-symbol interference (ISI), which significantly impacts the bit error rate (BER) performance of FTN multiple input multiple output (MIMO) optical wireless communication (OWC) system. To address this issue, an end-to-end (E2E) FTN-MIMO OWC system utilizing a long short-term memory (LSTM) autoencoder has been proposed. This approach effectively eliminates the influence of ISI and achieves complete signal recovery. Simulation results demonstrate that our proposal improves BER performance to different degrees under various conditions compared to conventional systems using maximum likelihood (ML) mothed. Specifically, with acceleration factors of 0.9 and 0.8, our proposal demonstrates an improvement in BER performance by 1.6dB and 2.2dB respectively, effectively mitigating the complex mixed ISI caused by FTN shaping and atmospheric turbulence channel.