Artificial Neural Network (ANN)-Aided Signal Demodulation in a SiPM-Based VLC System

Abstract

This paper describes a new type of data-driven signal demodulation technique using an artificial neural network (ANN) in a silicon photomultiplier (SiPM) based visible light communication (VLC) system. SiPMs can potentially be used to create the most sensitive optical receiver in VLC since it contains a large array of microcells and each microcell is capable of detecting individual photons. However, after each photon is detected, the associated microcell needs a period to recover. This means that the photon counting rate is not linearly related to the intensity of the light received by the SiPM and this effect causes a unique form of signal distortion when the transmission data rate becomes high. In this paper, an ANN aided data detection method is studied when the received signal is influenced by the SiPM nonlinearity. Both a single neuron based linear classifier and an ANN based non-linear classifier are introduced and explained in detail. The bit error rate (BER) results suggest that the ANN based receiver can significantly reduce the negative impacts caused by the SiPM nonlinearity for a wide range of irradiance levels.