Enhanced Performance of PAM7 MISO Underwater VLC System Utilizing Machine Learning Algorithm Based on DBSCAN

After visible light communication drawing increasing attention, underwater visible light communication (UVLC) has attracted more interest in the research community nowadays. As multiple input single output (MISO) is getting increasingly widely used to improve the transmission speed in UVLC system, the unbalance between multiple transmitters’ power is still a common phenomenon, which leads to the unequal spacing between each adjacent level and damages the system performance. In this paper, we study and analyze the unbalance between the two transmitters. Compared to a traditional hard decision, a density-based spatial clustering of applications with noise (DBSCAN) of a machine learning method is employed to get the actual center of each cluster and distinguish each level of PAM7 signals. In this way, a new decision curve substitutes traditional standard straight line as a constellation discrimination method. The experimental results show that up to 1.22 Gb/s over 1.2 m underwater visible light transmission can be achieved by using DBSCAN for PAM7 MISO signals. The measured bit error rate is well under the hard decision-forward error correction threshold of 3.8 × 10
⁻³
.