Experimental evaluation of an image sensor-based VLC method using object detection DNN for monochromatic color images

There is VLC (Visible Light Communication) technology that modulates and superimposes data signals on a signage image so that they are invisible to human eyes, and then receives data signals by capturing them with an image sensor, such as a smartphone camera. We have proposed a method that simultaneously processes the data signal detection and demodulation by applying object detection DNN (Deep Neural Network), and have demonstrated its effectiveness using simulated received images. In this paper, we experimentally evaluate the performance of the proposed object detection DNN-based signal detection and demodulation method. The 24 monochromatic color images based on the Macbeth color checker chart are used as the displayed signage image, and we clarify the effectiveness of the proposed method against to the displayed color.