K-Means++ Algorithm Based Detection of QPSK-PD-NOMA Signal over Free Space Optical Channel

In this work, we numerically estimate the performance of the conventional K-means clustering algorithm and K-means++ clustering algorithm based successive interference cancellation (SIC) for power division non-orthogonal multiple access (PD-NOMA) non-return to zero (NRZ) quadrature phase-shift keying (QPSK) signal over free-space optical (FSO) channel. Results obtained from numerical simulation show that the BER performance of K-means based SIC system is not consistent due to random selection of initial cluster position, whereas the BER results obtained by SIC using K-means++ initialization algorithm offers better performance. The OSNR penalty with respect to B2B for the near user (far user) offered by K-means++ assisted systems over FSO channel having $C_n^2$ values 5e−17, 5e − 16, and 5e − 15 are 0.2(0.2) dB,0.4(0.5)dB, and 2(2.−6)dB respectively. Results also reveal that the K-means++ initialization-based algorithm required a lesser number of samples compared to K-means algorithm to get a stable centroid, which is required for implementing the SIC.