Machine Learning-Assisted Receiver Design for Visible Light Communication Systems

Visible Light Communication (VLC) Systems have been attracted considerable attention from many researchers, papers, and projects due to their energy consumption efficiency. Hence, this paper focuses on VLC-based receiver scheme that is designed by using a machine learning algorithm to provide a sustainable demodulator structure with regardless to received signal format. To provide data transmission, the communication system uses SPAM (Superposed Pulse Amplitude Modulation) scheme which requires the use of threshold values to detect data bits. Hence, the receiver unit consists of a K-Nearest Neighbors (KNN) algorithm-based method to determine variable threshold value by using a training data set. An architecture has been enhanced to implement in Field Programmable Gate Arrays (FPGA) board. To integrate the machine learning based system into SPAM receiver model, a mathematical framework is improved to obtain the digital design techniques. Therefore, both a theoretical analysis and a digital design-based receiver system have been proposed in the paper. Addition to these, an Additive White Gaussian Noise (AWGN) channel model, which is designed by using Linear Feedback Shift Register (LFSR) is improved to observe performance of the Machine Learning based System related to the difference between transmitted and received bits.