Research On Channel Estimation And Signal Detection In OTFS

Orthogonal time frequency space(OTFS)modulation is a new emerging modulation scheme that aims to achieve reliable communication in high mobility scenarios.The basic idea of OTFS is to modulate the transmission symbols in the delay-Doppler domain.In this domain,each symbol in a data frame experiences the same almost constant fading,thus having better performance than existing modulation schemes such as orthogonal frequency division multiplexing(OFDM).At the same time,the OTFS system is equivalent to adding a pre-coding module and a post-decoding module to the OFDM system,so compatibility with existing communication systems can be achieved.The significant advantages of OTFS system have made it a study hotspot in recent years.Since channel estimation algorithms and signal detection algorithms have great impacts on the performance of OTFS system,this thesis focus on these two aspects.Firstly,the relevant theoretical basis of OTFS modulation is introduced in this thesis.Specifically,the content include the transmission characteristics of double-selective fading channels and its deterministic description,the basic principle of OTFS modulation and its equivalent channel matrix,which lay a theoretical foundation for subsequent analysis.Secondly,in order to further improve the accuracy of the channel estimation algorithm,this paper proposed a variational Bayesian inference(VBI)channel estimation algorithm.The channel estimation problem is modeled as a sparse signal recovery problem and a two-layer graphical model with the Laplacian scale mixture(LSM)prior utilized to model the sparse channel.Then,the VBI based algorithm is proposed to solve this problem.Simulation results are presented to show that the proposed algorithm can achieve better performance than other existing channel estimation algorithms.Finally,aiming at the problem of OTFS signal detection,a MMSE-DNN based signal detection algorithm is proposed.The received signal is preprocessed by the minimum mean square error(MMSE)detector and then used to train the deep neural network(DNN).The features of the training data are more easier for the DNN to learn after the MMSE detector,thereby improving the performance of the DNN detector.