Channel Gain Correlation and Prediction for WDM Underwater Wireless Optical Communication Systems

In order to reduce the receiver complexity and pilot overhead of wavelength division multiplexing (WDM) underwater wireless optical communication (UWOC) systems, we analyze the correlation and prediction of channel gain in this paper. We have chosen 16 wavelengths uniformly distributed from 450 nm to 600 nm, and obtained a dataset of WDM UWOC channel gains by Monte Carlo photon tracing method to obtain a. We analyze the channel gain correlation for WDM UWOC systems from various wavelengths. Numerical results suggest that the multi-wavelength channel gain has a strong correlation and the five largest main eigenvalues account for more than 99% of all eigenvalues. Based on the correlation analysis, we further propose a channel gain prediction algorithm using deep learning, which can predict channel gains only by some reference wavelengths. The proposed prediction algorithm significantly outperforms prior algorithms in terms of mean square error (MSE) under the same pilot overhead. Meanwhile, the proposed algorithm alleviates or solves the over fitting issue caused by too many reference wavelengths, which greatly reduces the receiver complexity and the pilot overhead.