Channel estimation of non-orthogonal multiple access systems based on L2-norm extreme learning machine

In this paper, we study non-orthogonal multiple access (NOMA) transmission system which is a promising technology in future 5G mobile communications. In this paper, we present an Extreme Learning Machine (ELM)-based channel estimation (CE) approach aiming at minimizing the effective bit-error-rate (BER) for a two-user NOMA downlink system. Channel estimation of a flat fading Rayleigh channel is carried out using two ELM-based algorithms. Simulation results of the proposed algorithm have been compared with other algorithms to justify the usefulness of the L 2 -norm ELM algorithm. A comparison table of the mean-square error (MSE) behaviour in a variable signal-to-noise ratio (SNR) condition has been included for the proposed NOMA scheme. Improved spectral efficiency (SE) and energy efficiency (EE) values can be observed for the proposed L 2 -norm ELM, and thus, it is proved that the ELM and L 2 -norm ELM-based NOMA schemes achieve a large sum capacity than other existing algorithms. In addition, a practical IEEE 802.11 g standard indoor channel has been modelled for NOMA system with the proposed L 2 -norm-based adaptive ELM channel estimator to evaluate the overall throughput performance.