DLSBL-OTFS：动态先验型SBL的OTFS信道估计方法

打开文本图片集

关键词：正交时频空间；DLSBL；稀疏贝叶斯学习；长短期记忆网络

中图分类号：TN929.5 文献标志码：A 文章编号：1001-3695（2026）01-027-0227-07

doi：10.19734/j.issn.1001-3695.2025.06.0177

DLSBL-OTFS ：dynamic prior-based SBL channel estimation method for OTFS

Zheng Juanyi，Wei Tiant （SchoolofCommuicationsndIfationEnginering，Xi’nUniersityofPostsndTelecomicatios，Xi'anhina）

Abstract：This paper proposeda DLSBL channel estimation method toaddresstheproblems of slow convergenceand poor generalizationinchannel estimation fororthogonal time-frequency-space（OTFS）systems.The method firstly employeda LSTM network tolearn and predictthe dynamic statistical characteristics of thechannel inthedelay-Dopper（DD）domain， generating accurateand time-varying sparsepriorinformation.Thisdynamic priorwas thenused toinitializethesparse Bayesian learning （SBL）algorithmfor chanel estimation，which solved the parameter selection problem in time-varying channels andffectivelysuppresses fractional Dopler interferenceandnoise.Simulationresultsshowthatthis methodsignificantlyimproves bit erorrate（BER）andnormalized meansquare eror（NMSE）compared toconventionalalgorithms.Theproposed method demonstratessuperiorobustnessinlowsignal-to-noiseratioandhigh-mobilityscenarios，andprovidesamoreefficient and accurate channel estimation solution for high-mobility wireless communication systems.

Key words：orthogonal time-frequency space（OTFS）； dynamic prior-based sparse Bayesian learning（DLSBL）；sparse Bayesian learning（SBL）；long short-term memory（LSTM）

0 引言

随着无线通信技术的快速发展，6G无线通信网络正逐步由传统的地面通信网络向空天地海一体化网络演进，以满足未来通信系统对更高数据速率、更低时延以及更高系统可靠性需求[1]。（剩余17539字）