计算机科学
高斯过程
人工智能
高斯分布
物理
量子力学
作者
Xufan Zhang,Lixia Xiao,Shuo Li,Qijiang Yuan,Luping Xiang,Tao Jiang
标识
DOI:10.1109/lcomm.2022.3206414
摘要
Orthogonal time frequency space (OTFS) modulation, which converts the fast time-varying channel in the time-frequency domain into a time-invariant channel in the delay-Doppler(DD) domain, can mitigate the fading caused by the dynamic change of the channel. Nevertheless, due to the influence of the Doppler effect, it is challenging for traditional algorithms to acquire the desired detection results. To this end, this letter proposes a Gaussian approximate message passing (GAMP) aided model-driven signal detection algorithm. It first expands the existing GAMP algorithm into a corresponding deep neural network. Then, a unique set of trainable parameters, including one damping factor and two stepping factors, are constructed to improve the algorithm’s convergence. Simulation results demonstrate that the designed model-driven signal detection algorithm outperforms the original GAMP by about 2 dB at the bit error ratio (BER) of$1\times 10^{-4}$.
科研通智能强力驱动
Strongly Powered by AbleSci AI