Analysis of a Multifarious User: Semantic-Aided Error Reduction

In this paper, we investigate an adaptive transmission (AT) design to facilitate the analysis of a multifari-ous user (MfU) that can operate in conventional mode, dual communication mode, or semantic mode. The AT enables the user to seamlessly switch between these modes. A haptics setup is employed to generate multidimensional data. Meanwhile, a metric, termed semantic class match, is introduced to quantify the performance across different modes. In the dual communication mode, power-domain non-orthogonal multiple access is utilized to transmit both semantic and conventional data. To enhance the performance in dual communication mode, two novel semantic-aided error reduction (SAER) techniques, namely the deep learning method and the average method are introduced. These techniques effectively reduce the mean square error (MSE) of conventional data. The deep learning method demonstrates superior MSE reduction capability. Furthermore, the semantic mode of communication is found to be very robust in low signal-to-noise ratio regimes. Numerical results highlight the significance of AT and SAER techniques in improving system performance for an MfU.