Deep Learning-Based Codebook Design for Code-Domain Non-Orthogonal Multiple Access: Approaching Single-User Bit-Error Rate Performance

A general form of codebook design for code-domain non-orthogonal multiple access (CD-NOMA) can be considered equivalent to an autoencoder (AE)-based constellation multi-user multidimensional modulation (MU-MDM). Due a constrained space optimal constellation, e.g., fixed resource mapping and equal power allocation all codebooks, however, existing AE architectures produce constellations with suboptimal bit-error-rate (BER) performance. Accordingly, we propose new architecture MU-MDM underlying training methodology joint optimization bit-to-symbol mapping, aiming at approaching the BER performance single-user MDM (SU-MDM) model same spectral efficiency. The core proposed is dense combined novel layer that normalizes sum user across entire resources. This globalizes domain by enabling flexible allocation. Furthermore, it allows AE-based approach global CD-NOMA. Extensive simulation results demonstrate outperforms CD-NOMA designs while achieved SU-MDM within 0.3dB over additive white Gaussian noise channel.