Semantic Communications for Speech Signals

We consider a semantic communication system for speech signals, named SCS. Motivated by the breakthroughs in deep learning, we explore the speech semantic to recover semantic meaning of the speech at the receiver, which aims to minimize the speech semantic error rather than the bit-error rate or symbol-error rate in traditional communications. Particularly, based on the attention mechanism squeeze-and-excitation (SE) networks, we design the transceiver as an end-to-end (E2E) system, which extracts and learns the essential speech information. Furthermore, in order to facilitate the E2E speech semantic communication system to work well on various practical communication scenarios, we find a model yielding good performance when coping with various channel environments without retraining process. The simulation results demonstrate that our proposed SCS is more robust to channel variations and outperforms traditional communication systems, especially in the low signal-to-noise (SNR) regime.