Semantic-Importance-Aware NB-PC Encoding and Modulation for Reliable Transmission Scheme

Abstract

In large-scale real-time multimedia transmission, different data segments have highly uneven semantic value. Traditional transmission treats all data equally, which leads to a mismatch between channel resources and critical semantic needs. To address this issue, this paper proposes a semantic-importance-aware non-binary polar coding (NB-PC) method. It adopts a two-level semantic importance evaluation mechanism, including class-level and channel-level analysis, and incorporates task-oriented importance weights to prioritize semantic content. High-priority semantic information is then mapped to highly reliable non-binary polarized subchannels for protected transmission. In addition, a semantic-guided Monte Carlo construction and a power-reallocation multilevel modulation scheme are designed. We theoretically prove that the proposed method enables the joint allocation of quantization and channel resources based on semantic importance. Simulation results show that the proposed method preserves critical semantic information under limited bandwidth and achieves a balanced tradeoff between compression efficiency and semantic fidelity.