Semantic Communication Systems for Speech Transmission

Semantic communications could improve the transmission efficiency significantly by exploring semantic information. In this paper, we make an effort to recover transmitted speech signals in communication systems, which minimizes error at level rather than bit or symbol level. Particularly, design a deep learning (DL)-enabled system for signals, named DeepSC-S. order recovery accuracy of especially essential information, DeepSC-S is developed based on attention mechanism utilizing squeeze-and-excitation (SE) network. The motivation behind identify information providing higher weights them when training neural Moreover, facilitate proposed dynamic channel environments, find general model cope with various conditions without retraining. Furthermore, investigate telephone systems as well multimedia verify adaptation practice. simulation results demonstrate that our outperforms traditional both cases terms metrics, such signal-to-distortion ration and perceptual evaluation distortion. Besides, more robust variations, low signal-to-noise (SNR) regime.