Experimental Demonstration of Continuous Sensor Data Monitoring Using Neural Network-Based Optical Camera Communications

Abstract

This technical paper presents an optical camera communication system by using data as input collected from a sensor that continuously garners humidity and temperature information from the environment. A small light-emitting diode (LED) and an LED array are used as a transmitter and a low-speed rolling-shutter camera as a receiver. A modulation scheme standardized in IEEE 802.15.7, referred to as variable pulse-width modulation, is used to encode the data bits. The proposed scheme is flicker-free in different frequencies and applicable in both static and mobile scenarios. A neural network is designed for LED detection and improving the bit-error-rate in mobile scenarios. A new method based on the region-wise comparison and an existing method is used to remove the interference and noise generated from the neighboring light sources, respectively. The decoding procedure is performed and analyzed in Python 3.7. A data rate of 1.02 kbps is achieved using the single LED, which is further augmented to 64 kbps by using an $8\times 8$ LED dot matrix.