深度学习在无线传输物理层的应用与实现

Application and Implementation of Deep Learning in Wireless Transmission Physical Layer

新型通信业务大量涌现和终端接入设备的急剧增长,导致信道建模复杂度显著提升,传统通信算法难以满足实时且精确地进行信号处理的要求,而深度学习(Deep Learning, DL)技术凭借其强大的模型学习能力、结构简单且运算速度较高的特点,成为无线物理层传输研究的主流方向。本文首先介绍了基于DL技术的三种经典神经网络,随后对DL技术在无线传输物理层模块如帧同步、编码器、检测器以及对整个接收机端到端替代的应用成果进行了总结和说明。

The explosive emergence of new communication scenario and the rapid growth of terminal ac-cess equipment have made channel modeling difficult, and traditional communication algo-rithms have difficulty meeting the requirements for real-time and accurate signal processing. Deep learning has the characteristics of strong model learning ability, simple structure and high operation speed, so it has become the mainstream direction of wireless physical layer transmis-sion research. This paper first introduces three classic neural networks based on deep learning, and then summarizes and explains the application results of deep learning in wireless transmis-sion physical layer modules such as frame synchronization, encoder, detector, and end-to-end replacement of the entire receiver.