基于轴向注意力机制的动态信道环境适配方法OA

To meet the requirements of 6G intelligent air interfaces for high robustness,low complexity,and environmental adaptability of receiver algorithms,this paper proposes an axial-attention-based channel adaptation method.The proposed method employs a low-complexity axial attention backbone to model long-range dependencies along the temporal and frequency axes,thereby enhancing the extraction of time-frequency correlated features from the two-dimensional OFDM resource grid.Meanwhile,a lightweight time-frequency adapter module is introduced to perform structured modulation of intermediate features through joint gating across the channel,temporal,and frequency dimensions,improving the model's rapid adaptation capability in dynamic channel environments.Experimental results demonstrate that the proposed method achieves superior block error rate(BLER)performance under different mobility conditions.In dynamic delay-spread scenarios,the parameter-efficient online update strategy effectively mitigates performance degradation with low update complexity and exhibits more robust online adaptation capability.These results validate the effectiveness of combining long-range time-frequency dependency modeling with lightweight structured adaptation,providing a feasible approach to improving the robustness of OFDM systems in dynamic environments.