A Theoretical Framework for Semantic Communication Systems in Low-Bandwidth Environments of Rural Nigeria

Semantic communication paradigms that transmit meaning rather than bits represent a transformative departure from Shannon's classical communication theory, and their theoretical performance characterisation in the severe bandwidth-limited and high-interference environments of rural Nigerian telecommunications (characterised by 2G EDGE connections at 150 kbps and below) represents an important original research question. This study developed a theoretical framework for semantic communication systems optimised for low-bandwidth rural Nigerian network conditions. The framework development employed a rigorous information-theoretic methodology, establishing new theoretical bounds on semantic transmission efficiency under AWGN and Rayleigh fading channels parameterised by measured rural Nigerian channel conditions. An original Semantic Rate-Distortion Theory was derived that incorporates task-specific distortion metrics beyond conventional MSE, enabling analysis of semantic accuracy preservation under extreme compression. A novel deep joint source-channel coding architecture (DeepJSCC-Rural) was designed with a transformer-based semantic encoder optimised for constrained bandwidth, implementing learned source-channel code rate adaptation based on real-time channel quality feedback. DeepJSCC-Rural was evaluated for image transmission (CIFAR-10 and field images from Nigerian agricultural extension use cases), achieving 14.3 dB better PSNR than JPEG with BPG compression at equivalent channel bandwidth usage under 2G EDGE channel conditions. A theoretical asymptotic analysis proved that the proposed architecture approaches the derived semantic rate-distortion bound within 1.8 bits per symbol under Rayleigh fading. The study constitutes an original contribution to semantic communication theory with direct applicability to rural Nigerian ICT access improvement.

Keywords: semantic communication, deep joint source-channel coding, rural Nigeria, rate-distortion theory, low-bandwidth