Research And Optimization Of Imaging Communication System Performance

Recently,as the development of the mobile Internet and Internet of Things,mobile traffic data is growing explosively.To achieve high capacity,ubiquitous connectiv-ity,high reliability and low latency,communication technologies in the high frequency band emerge as key solutions for the next generation mobile communication systems.The visible light communication(VLC),commonly known as LiFi(Light Fidelity),has many attractive advantages,including wide spectrum,no electromagnetic interference,high energy efficiency and easy implementation.It shows a great potential for deploy-ment in a spectrum-efficient and energy-efficient ultra-dense heterogeneous network.As a new form of VLC,optical imaging communication(OIC)has various appealing features and allows easier implementation.It employs an imaging sensor as a receiver,which is a natural optical MIMO,multi-color,and anti-interference receiver with a wide field-of-view(FOV).This motivates study on OIC for future high capacity optical wire-less communication networks.There have been various research works on the OIC modulations,synchronization and system implementation.But it's important to build a unified system model and evaluate analytical performance of an OIC system in order for the subsequent research in efficient signaling design,optimal signal detection and estimation,and channel coding.Based on the OIC system characteristics,we propose a unified OIC system model and further derive the capacity under different system settings.Meanwhile,we verify the proposed channel model and analytical capacity results via experiments.Accordingly,the main research contents and contributions of this dissertation are as follows:(1)We build a unified OIC communication model,consisting of the signal model,noise model and statistical channel model.Specially,we present the feasible space of nonnegative signals containing a set of multi-dimensional orthogonal basis functions,average power constrained and peak power constrained feasible space,all under the lighting constraints.We also build a statistical channel model,where the source radia-tion pattern,optical subsystem response function,pixel sensor response function,and pointing error model are considered.Meanwhile,we analyze the noise sources in the OIC imaging receiver based on the physical characteristics and manufacturing process of the optoelectronic device,and finally obtain a unified OIC communication model.(2)Assuming perfect channel state information at the transmitter(CSIT)is avail-able,the single-input single-output(SISO)imaging channel can be described as a mixed signal-dependent Gaussian noise(M-SDGN)channel.For such an M-SDGN channel,we prove the inapplicability of the entropy power inequality method,the input-output differential entropy method,the dual capacity method,and the sphere packing method in deriving the channel capacity.Then,we show that the capacity-achieving distribu-tion for the M-SDGN channel is unique and discrete with finite number of mass points,and also provide the sufficient and necessary conditions for the optimal input distribu-tion.Given those conditions,we prove that x = 0 is a capacity achieving mass point.Applying the Blahut-Arimoto iterative algorithm,we propose a low complexity search algorithm for the optimal discrete input distribution.(3)Assuming CSIT is available,we form the multi-input multi-output(MIMO)channel based on the multiple optical sources into the MIMO-SDGN channel model.Through analysis of such a channel,we find that the optimal source distribution achiev-ing the sum-capacity is no longer discrete.Afterwards,we transform the MIMO-SDGN channel into a set of parallel channels with bounded inputs and SDGN channels by ap-plying channel inversion,QR decomposition,and a DC-offset singular-value decom-position(SVD)scheme.We then present the upper and lower capacity bounds for the transformed parallel optical channels and propose a simple power allocation algorithm to achieve the sum capacity bound.According to the simulation results,it's demon-strated that the gap between the upper and lower bounds at high signal-to-noise ratio(SNR)is small,indicating the tightness of the bounds.It is further proved that the de-rived capacity bounds based on channel inversion and QR decomposition are higher than that based on the DC-SVD decoupling scheme.(4)We build an OIC experimental platform,and analyze the OIC noise character-istics.We compare the achieved data rate of the experimental platform with the derived capacity bounds.We verify experimentally that the OIC noise is white Gaussian,and the output noise variance is a quadratic function of the input signal,validating the proposed M-SDGN channel model.In addition,we measure the achieved data rate for the opti-mally designed SI SO imaging system and MIMO imaging system,and compare with the corresponding analytical capacities.The results reveal that there remain about a 2 bit/s/Hz?3 bit/s/Hz gap between the achieved data rate and capacity bounds for both the pulse amplitude modulation(PAM)-SISO imaging system and the color intensity modulation(CIM)-MIMO imaging system.