Development Of Semantic Communication Platform Based On USRP And Research On Semantic Coding Algorithm

In the era of big data,the increasingly scarce spectrum resources challenge the future wireless communication system with larger capacity.In order to meet the demand of efficient information transmission between intelligent communication entities,semantic communication aiming at "meaning" has become a research hotspot in recent years.The research on semantic communication mainly focuses on the representation and measurement of semantic information,encoding of semantic information and transmission of semantic information.Aiming at massive data interaction scenarios,how to realize semantic information transmission in wireless communication system is of great significance for making full use of spectrum resources and further improving communication efficiency.The topic of this paper is from the Major research program of National Natural Science Foundation of China《Semantic Driven Industrial Internet Native Intelligent Simple Organization Theory》(Project No.:92067202).In this paper,semantic information transmission is realized on 5G link-level communication experimental platform based on USRP by designing semantic communication platform architecture oriented to intelligent task.Furthermore,a semantic coding algorithm based on semantic importance is proposed to solve the problem of spectrum resource shortage,which can reduce the bandwidth requirement while ensuring the task performance.Relying on this platform,he presided over the completion of the "2022 Graduate Innovation and Entrepreneurship Project of Beijing University of Posts and Telecommunications",and participated in the "Advanced Information Network Beijing Laboratory Scientific Research Achievements Exhibition".The main work completed in this paper is as follows:1.Research and review the key technologies of semantic communication.Firstly,the basic principle of semantic communication technology and the architecture of semantic communication system are summarized.Secondly,the key technologies of semantic information representation and measurement,semantic information encoding and semantic information transmission are studied in depth.Then,it summarizes the semantic communication experiment architecture and the research basis of 5G link-level communication experiment platform,which provides the basis for the design and implementation of semantic communication platform based on USRP.Finally,the current problems and challenges are analyzed.2.Design and implement semantic communication platform based on USRP.Firstly,in view of the practical problems of semantic transmission in communication system,the requirements of semantic communication platform are analyzed from three aspects:semantic level,technical level and layer indirect interface.Then design and implement the platform according to the above functional requirements.The semantic layer realizes the extraction and analysis of semantic information of source data.The technology layer realizes the transmission and reception of physical signals of 5G communication link by controlling hardware equipment;The layer indirect interface realizes the mapping between semantic inf-ormation at the semantic layer and 5G physical signals at the technical layer.Finally,taking the task of surface defect classification of hot rolled strip steel in industrial Internet scene as an example,the functional test and performance analysis of the platform are carried out.The experimental results show that,compared with the traditional communication mode,the bandwidth utilization rate of the platform in processing intelligent tasks is increased by 10 times,the end-to-end task delay is reduced by 46.8%,and it has the advantages of strong anti-noise ability.It not only provides strong support for semantic communication research,but also provides an important reference prototype for the implementation of semantic communication system architecture.3.Study semantic coding algorithm based on semantic importance.Firstly,according to the requirement of combining wireless communication system with intelligent task processing,a semantic communication system model oriented to intelligent task is established.Secondly,a semantic coding algorithm based on semantic importance is proposed to limit the task bandwidth requirement.By aggregating channel features and adaptively calibrating feature responses,the algorithm obtains the importance weight of feature maps to task results,and screens feature maps with large weight during transmission to achieve semantic compression.Finally,the proposed algorithm is tested and analyzed on the semantic communication platform.Experimental results show that the proposed algorithm can compress 80%of the feature data and reduce the task delay by 41.9%on the premise of ensuring the task performance.It is proved that it is suitable for resource-constrained and delay-sensitive scenarios,and has universality to different classification networks.