基于深度神经网络的低轨卫星信道估计动态映射策略OA

In low-earth-orbit(LEO)satellite communication systems,addressing the poor performance of traditional chan-nel estimation algorithms under dynamic delay variations and low signal-to-noise ratio(SNR)caused by high-speed satellite-terrestrial motion,a deep neural network(DNN)-based channel estimation algorithm combined with sub-block merging of channel responses under an orthogonal frequency-division multiplexing(OFDM)framework was proposed in this paper.The reliability and stability of data transmission in non-terrestrial network(NTN)were enhanced in this method.The algorithm divided the frequency-domain pilot channel response matrix into sub-blocks for averaging and em-ployed a DNN to adaptively determine the optimal sub-block merging number under different channel conditions with varying delay and noise.The bit error rate(BER)performance of this algorithm was compared through simulations with the existing channel estimation methods under dynamically randomized delays.Simulation results demonstrate that the proposed algorithm achieves lower BER at identical SNR levels than traditional methods,exhibiting stronger robustness.Compared with the existing neural network-based channel estimation algorithms,it adapts better to larger delays and dy-namic delay variations,showing superior generalization capability.