Abstract
Due to an increase in mobile and computer devices, internet traffic is increasing remarkably every year. To solve this issue, a lot of research is being conducted in the field of Visible Light Communication (VLC). As it can revolutionize the existing radio frequency-based communication system by upgrading the speed and security. For such communication systems, signal classification is one of the most important tasks. In this project, we propose deep learning-based algorithms for signal classification in the VLC system. Implementations of deep learning algorithms require a large amount of data. For that, an end to end VLC system is implemented with two Arduino devices and four different modulation algorithms. In the VLC system, a white LED is used to transmit the data for the transmitter Arduino, and a photoresistor is used to receive the data for the receiver Arduino. The four modulations implemented are On-Off Keying (OOK), Quadrature Phase-Shift Keying (QPSK), Pulse-Width Modulation (PWM), and Pulse Position Modulation (PPM). The data is collected from various distances between the transmitter and the receiver, starting from the less to the greater distances between the two Arduinos. In this project, we use Convolutional Neural Network (CNN) to classify the modulated signals. We convert the collected signal data into images to feed them as an input to the CNN algorithm. For the greater distances, we use the Conditional Generative Adversarial Network (CGAN) to augment the signal dataset and again classify them using the CNN to make the classifier more accurate.