An efficient adaptive thresholding scheme for signal decoding in NLOS VLC systems

Visible Light Communication (VLC) has been evolved as a promising technique for indoor communications in the last years, exhibiting several merits compared to existing Radio Frequency (RF) communications. In the case of LED-to-Camera systems, the demodulation of the transmitted signal is achieved by exploiting the rolling shutter effect in the frames captured by CMOS sensors. The processing of each frame aims at the correct identification of the bright and dark stripes in the region of interest. Furthermore, the accurate determination of the width of each stripe has a significant influence in the performance of the system, when the Manchester encoding of the signal is used. However, the exact width is usually ambiguous, due to the severe fluctuation in the image intensity. In this work, we present an adaptive thresholding technique for the classification of the stripes in one-symbol or two-symbols categories, for the demodulation of the transmitted signal. The proposed technique was evaluated and compared with a constant threshold-based scheme in a non line-of-sight (NLOS) system, and the results indicate that our method is more efficient.