Multi-cell MIMO Semantic Communication based Distributed Training.

Driven by deep learning, natural language processing(NLP) has achieved great success in analyzing and understanding large volumes of text. As a result, a large number of communication transmission methods based on NLP have emerged, which can identify the semantic information behind data and compress the amount of transmitted data to achieve better transmission performance. However, applying natural language processing to practical scenarios still faces numerous challenges, such as user interference among multiple cells and limited computing power at base stations. To address these issues, we propose a multi-cell multiple input multiple output(MIMO) semantic communication transmission method. We also consider the limited computing power at base stations and introduce a distributed training method. Each base station trains independently and uploads its network weights to a central server for fitting. The fitted weights are then used as the initialization parameters for the next round of training. Simulation results show that our proposed transmission method is robust under different signal-to-noise ratios (SNR). Additionally, the distributed training at each base station can better understand the semantic information from other knowledge bases.