The Inner And Outer Bound Of The Air-ground Joint Information Perception CEO System

The Distributed Source Coding(DSC)theory was introduced in 1973 and has provided key guidance for solving many coding and transmission technical problems.However,the current distributed source coding theory mostly considers the compression reconstruction of scalar source or vector source,which makes the system unsuitable for some complicated scenarios,so it still has many open problems.For example,when the source to be reconstructed by the system is not all scalar or vector,some sources are affected by large-order noise,and the number of sources is not always smaller than the number of encoders,how to determine the rate-distortion region of the system.These problems have greatly improved the applicability of the system in certain scenes,such as air-ground joint information perception scene.The solution of such problems helps to analyze the problem of source compression limit in different scenes from the information theory level.This dissertation proposes a new distributed source coding system,which can not only be applied to the air-ground joint information perception scene,but also provides some guidance for the research of other scenes data compression problems.In this dissertation,based on the existing distributed source coding theory,a new CEO(Chief Executive Officer)source coding system,which is called air-ground joint information perception CEO system,is introduced.The system consists of a vector source,multiple scalar sources,multiple encoders,and a joint decoder.The encoder in the system encodes the source of noise version,and transmits the encoded result to the joint decoder for decoding.The decoder attempts to decode and reconstruct the original vector source.In this system,the noise versions of saclar sources are used as the side information at the decoder.The system allows the number of sources to be greater than the number of encoders,while the intensity of the noise received by each dimension of the vector source will be greater than that received by each scalar source.In order to further describe the system,this dissertation takes the scene of the vehicle network intersection in the air-ground joint information perception as an example,the aerial observation node in this example scene perceives aerial and fuzzy information,the ground observation nodes in this example scene perceive local and arrcurate information.According to the distributed source coding system,the outer and inner bound of rate-distortion are solved.The derivation of the outer bound is based on a modified Entropy Power Inequality(EPI).The outer bound shows that the lossy source compression can be achieved when the rates are not lower than the required rates under the given distortion.The inner bound is solved based on Slepian-Wolf coding system and Berger-Tung coding system,from pre-encoding,encoding scheme,decoding scheme,and error estimation.The inner bound indicates that there is at least one source coding scheme in theory,which makes the decoder can reconstruct the estimate of the source with a probability of asymptotic to 1,and the expected distortion of the source and source estimates is close to the given distortion.In order to illustrate the performance advantages of the air-ground joint information perception CEO system,we compare and analyze the conclusions of the Gaussian CEO system studied by Oohama et al.It is shows that the outer bound of the rate-distortion region corresponding to the air-ground joint information perception CEO system are tighter to the rate-distortion region corresponding to the system established by Oohama et al.,that is,the new system can use fewer bits to describe the original source.In particular,when the vector source in the new system degenerates into a scalar source,and the number of sources changes,the air-ground joint information perception CEO system will degenerate into distributed source coding systems studied by Oohama et al.,and the result will be degraded to the results of Oohama et al.This explains the significance of the proposed distributed source coding system and the correctness of the conclusions obtained.