An Original Theoretical Investigation of Semantic Communications for Bandwidth-Constrained Telecommunications in Rural Nigeria

Semantic communication represents a paradigm shift from the Shannon information theory framework, transmitting the meaning and task-relevant content of messages rather than bit-level representations, offering the potential for dramatic reduction in bandwidth requirements that is particularly relevant for rural Nigerian communities served by severely capacity-constrained networks. This doctoral study conducts an original theoretical investigation of semantic communication systems for bandwidth-constrained rural telecommunications, developing a mathematical framework that unifies semantic compression theory with channel coding constraints relevant to Nigerian rural propagation environments. An original semantic compression bound is derived for image and speech modalities under a task-oriented communication framework, characterizing the minimum semantic information content required to support specified downstream task performance targets, including telemedicine image diagnosis accuracy and voice command recognition fidelity. The theoretical bound is derived using a rate-distortion approach in which distortion is measured in semantic task performance space rather than pixel or sample space. A deep learning-based semantic codec architecture implementing the derived theoretical framework is proposed and evaluated on a dataset of Nigerian language speech commands and medical images relevant to primary healthcare applications. The proposed system achieves equivalent task performance using 73% less bandwidth than conventional image compression at the distortion threshold relevant to telemedicine diagnosis tasks. Theoretical analysis of robustness to channel impairments characteristic of rural Nigerian 3G environments demonstrates graceful degradation in task performance under packet loss conditions. The doctoral study advances semantic communication theory by deriving the first task-oriented rate-distortion bounds for telemedicine and voice command applications. Keywords: semantic communication, bandwidth efficiency, rural Nigeria, rate-distortion theory, telemedicine.