融合深度学习与注意力机制的信道建模方法

A Channel Modeling Method Integrating Deep Learning and Attention Mechanism

有了有效解决高速运动场景下信道建模不准确的问题,提出融合深度学习与注意力机制的信道估计方法。该方法通过深度学习和注意力机制重构信道频率响应值以解决信道时变的误差补偿问题;在此基础上,为了有效应对通信环境的小范围动态变化,基于信道估计误差来动态调整信号采样间隔,解决信道小范围波动所导致计算资源浪费问题。仿真结果证明,该方法能为动态场景下自感知、自决策网络信道建模方法提供理论支撑。

To effectively solve the problem of inaccurate channel modeling in high-speed motion scenarios, a channel estimationmethod combining deep learning and attention mechanism is proposed. This method reconstructs the channel frequencyresponse value through deep learning and attention mechanism to solve the error compensation problem of channel timevaryingproblems. On this basis, in order to effectively respond to small-scale dynamic changes in the communicationenvironment, the signal sampling interval is dynamically adjusted based on channel estimation error to solve the problemof computing resource waste caused by small-scale fluctuations in the channel. The simulation results demonstrate that theproposed method can provide theoretical supports for the channel modeling methods of self-perception and self-decisionnetworks in dynamic scenarios.