Data-Driven Receiver for OTFS System with Deep Learning
Qingyu Li, Yi Gong, Fanke Meng, Zhan Xu
Abstract
Recently researches about receiver structures for orthogonal time-frequency space (OTFS) have been received widespread attention. Previous OTFS receiver algorithms are based on model-driven, which would lead to complex structures. Motivated by recent advances in data-driven receivers, this paper proposes a data-driven OTFS receiver with a deep neural network (DNN). We demonstrate that the proposed data-driven receiver for OTFS can be generalized to different high mobility scenarios. Specifically, this scheme combines the power of deep learning (DL), which is widely used in various fields. With DL, the proposed algorithm can achieve excellent robustness and strong generalization ability for channel parameters, which are ubiquitous challenges in the design of receiver algorithms. Through a good deal of numerical experiments, simulation results show that the proposed data-driven receiver based on DNN for OTFS can achieve superior performance than comparison methods.