Channel Characterization,Achievable Rate,and Signal Detection For Weak Light Communication

Optical wireless communication(OWC)can serve as a potential comple-mentry technique to conventional radio frequency(RF)communication,since it occupies a bandwidth from 380nm to 780nm,which can alleviate the bandwidth scarcity due to the wide usage of wireless terminals.Light emitting diode(LED)can not only be light source but also transmit information.Recent research on VLC has involved the modulation scheme,coding,multiple-input multiple-output techniques(MIMO),link modeling,transceiver architecture and so on.Most ex-isting works on VLC focus on the linear operating regime of the transmitter,with relatively strong ambient light and signal component at receiver side.However,the application of VLC may not be limited to strong signal and ambient light region.Other weak light communication scenarios can be considered,for example,the long-distance transmission in deep sea water,long-distance transmission in coal mines,the weak optical signals due to blockage,and non-line-of-sight(non-LOS)scattering communication,which exhibit the characterizations of weak signal com-ponents.Thus,this thesis will investigate the signal characterizations,detection and system performance of weak light communication under different background environments.This thesis investigates the signal model and characterization of weak otpical signals under weak illuminance,and also the signal characterization,signal detection under strong ambient light.Finally,this thesis introduces the sequential detection to the weak signal of scattering optical communication.This thesis provides several innovative studies on the optical signal character-ization,achievable rates and signal detection based on system experiments.The main contributions are summarized in the following.·Under weak illuminance,the signals converted by PMT show different char-acteristics,as the received power increases,the output signals exhibit from the initial discrete pulse signals to continuous waveform signals.This thesis proposes three signal detection approaches,the waveform-sampling detec-tion,the pulse-counting detection,and the combination of the two detection approaches.Moreover,this thesis discusses the three detection approaches under different conditions of signals intensity and sampling rate.· Under weak illuminance and weak signal power,significant nonlinearity can be observed from the receiver side in experiments.Based on the experimen-tal measurement,this thesis has characterized the LED nonlinear effect via a state transition model,and adopted hidden Markov model to characterize the transmitted signal.A Monte-Carlo method is proposed to compute the achievable transmission rate for such weak signal communication system,and a Viterbi algorithm is developed to detect the pulse signals.Experi-mental results show that the signal detection based on the state transition model can achieve lower detection error probability.· Under strong ambient light,the characterization of Avalanche Phot,odiode(APD)output signals is investigated-This thesis illustrates the saturation of APD due to strong ambient light.And the saturation of APD decreases the output signal amplitude at the receiver-side.Based on the signal char-acteristics,the thesis has investigated the achievable link transmission rate given the transmission power and distance.· Based on photon-counting receiver and PMT detector,the thesis adopted sequential detection approach to weak signal detection in scattering optical communication.-The arrivals of photons in photon-counting receiver satisfy the Poisson distribution.Based on the predefined miss detection and false alarm probabilities,the thesis compares the performance of sequential de-tection and fixed-sample-size detection.And the sequential detection requires less samples compared with fixed-sample-size detection,which is more efficient for detection.-For the PMT receiver,the thesis proposes the mixed-Gaussian signal model due to photoelectric conversion process.One-term approxima-tion and Gaussian approximation approaches are proposed to approx-imate the true mixed-Gaussian distribution,which substantially re-duces the computational complexity.This thesis also compares the performance of One-term approximation,Gaussian approximation ap-proaches and the mixed-Gaussian based detection.And the One-term approximation and Gaussian approximation approaches requires close even less number of samples than the mixed-Gaussian based detection.