NOMA VLC Systems and Neural Network Approach for Imperfect SIC

Abstract

Non Orthogonal Multiple Access (NOMA) technique in Visible Light Communications (VLC) enhances the performances like spectral efficiency, achievable data rate, fairness, outage probability, etc. NOMA uses superposition in power domain at the transmitter and Successive Interference Cancellation (SIC) at the receiver. SIC operation is expected to perform perfect cancellation to avoid errors in the received signal. In this paper, Neural Network (NN) methods are used to overcome imperfect SIC in a NOMA VLC system. The Signal to Noise Ratio (SNR), Bit Error Rate (BER), and bitrate performance of the NOMA VLC systems are analyzed using Convolution Neural Network (CNN), long short term memory (LSTM), and Deep Neural Network (DNN) algorithms. Simulation results shows that the NN methods outperforms the conventional NOMA VLC system to a perfect SIC. Considering SNR for the BER 10−4, CNN outperforms SIC by 5 dB, DNN by 2 dB and LSTM by 1.5 dB. Further, CNN also outperforms SIC, DNN, and LSTM based NOMA VLC systems for BER performance as a function of bitrate. Thus, NN-based receiver will be a better alternative for imperfect SIC.