Abstract:
Visible light communication (VLC) is an emerging technology aiming to use lighting equipment to convey data via modulating light intensity. With the running out spectrum for radio frequency wireless communication, VLC is gaining more at tention in the literature and its practical issues are investigated intensely. Two main challenges distinguish VLC from other communication systems: satisfying illumination requirements during data transmission, and the hard channel presented by the trans mitters’ (usually Light Emitting Diodes (LED)s) limited bandwidth and inter-symbol interference. IEEE 802.15.7 group has defined the required standards for VLC. The standard proposes the usage of run-length limited (RLL) codes to provide constant dimming levels while transmitting data. In this thesis, the VLC channel is investi gated and experimental measurements for frequency modeling of LEDs are conducted first. Two different approaches for performance improvement of VLC links are then proposed. In the first approach, an RLL code that guarantees a maximum run-length of 3 is proposed and experimentally tested. By guaranteeing the run-length limit, the code prevents the vanishing of bits due to the high-pass effect of equalizers. Conse quently, the bit error rate (BER) improvement is verified. In the second approach, a high-performance, robust deep learning-based end-to-end VLC system is proposed. The proposed model provides dimming support by modulating the data in zero-mean codewords that can be superimposed with arbitrary DC levels for dimming control. The model is then modified to compensate for the LEDs’ nonlinearity. Numerical re sults show that the proposed model can operate in two different channels with the same high performance.
