Signal demodulation method based on a temporal-convolution feature fusion network for an ultraviolet communication OOK-NOMA system.

In this paper, a demodulation method based on a temporal-convolutional feature fusion network (TCFFN) is proposed for the non-line-of-sight (NLOS) ultraviolet communication (UVC) system. The TCFFN extracts the temporal features and the local features of the signals, offering strong adaptability to inter-symbol interference (ISI) caused by channel scattering. By evaluating a single-user and dual-user UVC on-off keying non-orthogonal multiple access (OOK-NOMA) systems, the results demonstrate that the TCFFN demodulator supports the higher rate transmission of NLOS UVC system compared with the static threshold (ST) demodulator and the minimum mean square error (MMSE) equalizer. In the dual-user scenario, the reliable communication rate with TCFFN reaches 8 Mbps in both the coplanar and the first non-coplanar configuration, and 4 Mbps in the second non-coplanar configuration, while the system bit error rate (BER) fails to reach the forward error correction (FEC) with using neither ST nor MMSE.