Deep Learning Assisted Pre-equalization Scheme for Faster-than-Nyquist Optical Wireless Communications

Faster-than-Nyquist technology introduces additional inter-symbol interference when improving the data rate and spectrum utilization. A deep-learning based pre-equalization scheme is proposed to reduce the inter-symbol interference. The bit error rate performance under different acceleration factors is analyzed. Monte Carlo simulation results show that the bit error rate performance can be improved by 1dB, 5.1dB, 6. 0dB and 8. 6dB when the acceleration factor is 0.9, 0.8, 0.7 and 0.6, respectively.

I. Introduction

Optical wireless communication (OWC) has the advantages of high speed, large capacity, wide spectrum and flexible links [1–5]. Therefore, it becomes one of the main alternatives to the requirements of large-capacity, high-bit-rate and low-latency in the beyond 5G era. However, OWC is susceptible to the weather conditions, atmospheric turbulence and hardware speed limitations, which limits the data rate increase. On the other hand, as a non-orthogonal transmission technology, Faster-than-Nyquist (FTN) can send more data under the same bandwidth [6–8]. With the increasing demand of high data rate links, FTN technology is attractive in OWC communications to further improve its data rate [9–11]. Different from traditional methods that require more resources like time slot, spectrum bandwidth, and space to improve the data rate. FTN improves the spectrum efficiency by artificially compressing the symbol interval to transmit more symbols. As a result, FTN can achieve the ultimate capacity of signal power spectral density (PSD) [12]. Consequently, FTN was rediscovered as a promising technology and attracted widespread attention from industry and academia [13–20].