基于深度学习的MIMO系统联合信号检测与信道译码

为进一步提高多输入多输出(multiple-input multiple-output,MIMO)系统性能,研究了深度学习方法来联合解决MIMO系统信号检测与信道译码问题.通过将深度神经网络、自动编码器神经网络与传统MIMO通信系统的物理层架构进行有机融合,构建了基于神经网络的MIMO系统模型,可获取系统发射端的信息比特或码字及信道状态信息,采用了端对端的训练方式,使不同神经网络模型可学习系统收发端的信息比特与码字的映射关系,联合实现了MIMO系统信号检测和信道译码,同时具有较低的复杂度.仿真结果表明,相比一些传统检测和译码算法,所提方法具有较优的检测和译码性能.

Joint signal detection and channel decoding for MIMO systems based on deep learning

In order to further improve the performance of multiple input multiple output systems, a deep learning method is studied to jointly solve the problem of signal detection and channel decoding. Two neural networks, deep neural network and autoencoder neural network, are organically integrated with the physical layer architecture of the traditional multiple-input multiple-output (MIMO) communication systems to build MIMO systems model based on the neural network, which can obtain the information bits or codewords at the transmitter and channel state information. An end-to-end training mode is adopted, so that the different neural network models can learn the mapping relationship between the information bits and codewords on the transmitting and receiving end of the MIMO systems. Therefore, signal detection and channel decoding can be jointly realized with lower complexity. Simulation results show that the proposed method can outperform the performance of some traditional detection and decoding algorithms.