Performance Enhancement for Non-Light-of-Sight Optical Camera Communication Using RPSNN

Non-line-of-sight (NLOS) optical camera communication (OCC) offers a promising avenue for improving link availability, thanks to its resilience against link blockage. Nevertheless, the diverse reflection coefficients across various reflective points introduce background interference and inter-symbol interference (ISI). Thus, the pixel values in one specific column may suffer severe fluctuation, complicating signal demodulation. To mitigate the background interference and improve the transmission performance in NLOS OCC system, in this letter, row pixel sorting based neural network (RPSNN) is proposed and experimentally demonstrated. The results indicate that with a white and heterogeneous background under reflective illuminance of 70 lux, the RPSNN scheme attains a data rate of 8.16 kbps with an average bit error rate (BER) of 1.4x10(-3) and 5.4x10(-3) . In contrast, the conventional artificial neural network (ANN) without row pixel sorting only achieves a BER of 2.3x10(-2) in white background and even worse in heterogeneous background. These results prove that the proposed RPSNN scheme is a promising scheme for OCC system in NLOS scenarios.