Deep Semantic-Aware SCMA Codebook Learning for Semantic Communication Systems

Sparse code multiple access (SCMA) has emerged as a promising non-orthogonal multiple access (NOMA) technique for future wireless communications. However, conventional SCMA is inherently designed for bit-level transmission, which ignores semantic-level information and lacks semantic-aware codebook construction capabilities. To address this limitation, this paper proposes a semantic-aware SCMA codebook learning framework, named SCMA-SC, which jointly optimizes semantic representation and sparse codeword generation through an endtoend trainable architecture. By integrating semantic communications with SCMA, the proposed framework enables semanticaware overloaded transmission over limited wireless resources. Simulation results demonstrate that SCMA-SC outperforms both conventional SCMA and power-domain NOMA semantic communication systems in terms of semantic reconstruction quality and image transmission performance.