Performance analysis of coexistence of traditional communication system and emerging semantic communication system

Abstract

As a new paradigm, semantic communication (SC) is attracting attention as the next-generation communication. In order to develop semantic communication systems, the research trend is to adopt deep learning networks with the best performance in each field of deep learning. Deep-learning based SC systems are task-dependent because the deep networks employed are specialized to perform specific tasks. Therefore, SC systems cannot be used for users using applications that do not work with the deep networks exploited by the SC systems. Such users still should be served by a traditional communication (TC) system. Hence, in the near future, communication networks will face the coexistence of both SC system and TC system, and thus a base station (BS) has to simultaneously transmit symbol streams encoded by an SC system and a TC system via orthogonal frequency-division multiple access. Although the current works verify that SC systems outperform a TC system in a point-to-point communication, it is necessary to analyze whether the introduction of SC systems can also improve the network performance provided by a BS. In this paper, we investigate the impact of introducing the new emerging semantic communication on network performance by modeling signal-to-noise ratio of SC users served by SC systems and analyzing sum-rate maximization with a minimum required SNR constraint for SC users. Via extensive numerical results, changes in network performance with and without an SC system are discussed with varying the distances from a BS, the number of SC users served by an SC system, and the degree of SC performance.