Since indoor visible light communication (VLC) modulates the information into the light beams emitted from light-emitting diodes, the channel gain is often modeled as the Lambertian model. For different spatial locations of transmitters and receivers, the channel gain for VLC directly changes with several related geometrical parameters. The distinct geometrical feature makes the ergodic capacity of indoor VLC different from that of radio-frequency communication. However, the issue has not been studied currently. In this paper, we propose the lower bounds on the point-to-point capacity, which have simple expressions with respect to the geometrical parameters. Then, by analyzing indoor human mobility, two typical distributions of the geometrical parameters are considered. Correspondingly, we derive the lower bounds on the ergodic capacity, which are related to the variables of the geometrical feature and the distribution model. Furthermore, simulation results show that our lower bounds on the ergodic capacity are effective to reveal the geometrical feature of indoor VLC systems. Moreover, our proposed lower bounds are tight to the numerical upper bounds at high optical signal-to-noise rates, which are the main application zones of indoor VLC systems.

There is Poisson slot noise generated by the single photon avalanche diode (SPAD) and Gaussian thermal noise generated by the transimpedance amplifier, respectively, if an analog SPAD detector is applied in the visible light communication (VLC) system. In this letter, for the long distance SPAD-based underwater VLC system, a fast blind multiple-symbol detection (MSD) algorithm is developed under the pure Poisson noise model. And then, a blind MSD algorithm based on generalized maximum-likelihood rule under the Poisson-Gaussian (P-G) mixed noise model is proposed. A generalized Anscombe transform (GAT) is employed to make the P-G mixed noise model as an equivalent additive white Gaussian noise model, then an equivalent minimum Euclidean distance criterion is presented to formulate the optimization problem. Simulation results show that our blind MSD scheme can be identical to the performance under a GAT-based detector with perfect channel state information.

Multi-input multi-output (MIMO) holds great potential in the high-speed transmission of visible light communication (VLC). However, as for the current MIMO-VLC techniques, the error performance is improved by wasting spatial resources, which just drops the core advantage within an MIMO structure. Moreover, they cannot change over the channel, which greatly limits their application range. In this paper, a general design criterion of the channel-adaptive space-collaborative constellation (CASCC) is established for MIMO-VLC systems. Specifically, we not only consider the channel into our criterion, but also build relations among the light-emitting diodes to exploit the spatial resources instead of wasting. Under the criterion, a regular CASCC is particularly designed for $2\times 2$ MIMO-VLC in terms of the basic four-point constellation and the high-order constellation. Correspondingly, a fast maximum likelihood (ML) detection algorithm for CASCC is proposed. Simulation results show that our proposed CASCC obtains better error performance and wider application range compared with the conventional MIMO-VLC schemes. Moreover, the fast ML algorithm reduces the computational complexity without any performance loss.

According to our experimental observation, the effect of dead time caused by the quenching circuit of single photon avalanche diode (SPAD) devices could affect the system performance significantly. Hence, we propose a corresponding synchronization scheme and a fast blind union detection (FBUD) algorithm for the received photon-counting pulse waveform by considering the effect of dead time. To verify our proposed theory under the long distance turbulence fading channel, we demonstrate a long distance SPAD-based visible light communication system with a red light-emitting diode status indicator in the practical circuit board. The experimental results indicate that the proposed FBUD algorithm outperforms the conventional dead-time free Poisson multiple-symbol detection method under different conditions. Meanwhile, the bit-to-error ratio gaps of the two algorithms will become more and more larger with the increasing of the data rate.

Visible light communications (VLC) using multi-color light-emitting diodes (LEDs) can support simultaneous high-speed data rate and high-quality lighting. However, since the radiation spectrum of LEDs has a limited width, spectral overlapping will result in multi-color cross talk even when optical filters are applied at the receivers. Moreover, since LEDs are used for illumination and wireless data transmission in the meantime, both lighting quality and communication performance must be considered in VLC systems. In this paper, we consider a multiple-input-multiple-output with low-complexity linear minimum mean square error detection to collaboratively manage the cross talk by maximizing the minimum signal-to-interference-plus-noise ratio (SINR) subject to chromaticity constraint based on MacAdam ellipse, luminance constraint, and signal range constraint. A sub-optimal convex relaxation is proposed to attack the SINR maximization problem. Extensive simulations indicate that the proposed method provides very efficient solutions and outperforms the conventional wave-division multiplexing scheme under the illumination constraint.