Incorporating Importance Sampling in EM Learning for Sequence Detection in SPAD Underwater OWC

Abstract

In this paper, an importance sampling-based expectation-maximization (EMIS) algorithm is developed for sequence detection in single-photon avalanche diode underwater optical wireless communication (OWC) systems. To be more specific, the expectation-maximization (EM) algorithm in statistic learning provides a general framework for the sequence detection, and the importance sampling (IS) method is employed for evaluating the minimum mean-square error estimates required in the EM algorithm. Theoretical analysis indicates that the developed EMIS algorithm achieves near-optimal performance with low-complexity symbol-by-symbol detection. The simulation results verify the effectiveness of the proposed EMIS algorithm.