Abstract
This paper investigates a novel multi-user multiple access semantic communication system called semantic feature multiple access (SFMA). Our semantic system consists of one base station (BS) and multiple users. The users are clustered into several groups, each comprising two users. The BS first generates the semantic information of the source images users request via semantic encoders. Then, the BS combines the semantic information transmitted to the two users in a group into a single transmitted signal using their semantic features. The users will extract their images from the combined semantic information. To optimize their transmission rates, it is necessary to optimize the transmit power that the BS allocates to each user. This problem is formulated as an optimization problem whose goal is to maximize the sum rates of the system while meeting each user's minimum rate requirement. To solve this problem, we first use simulations to verify that the standard equation of signal-to-interference-plus-noise (SINR) ratio cannot capture the performance of our designed multi-user multiple access semantic communication system since it cannot capture data meaning transmission performance. We then introduce a weight parameter into the standard SINR equation to accurately capture the performance of our designed semantic system. However, since this weight parameter depends on the transmit power of the users, it will significantly increase the complexity of finding the optimal transmit power for each user. To this end, we propose a two-step solution. In the first step, we find the optimal power that should be allocated to each group. In the second step, we optimize the power that should be allocated to each user within a group. Simulation results show that our proposed method can improve the transmission rate by up to 18.6% and 37.1% respectively, compared to the fixed non-orthogonal multiple access (F-NOMA) and the orthogonal joint source-channel coding methods.