基于DRLNet的一种OFDM系统信道估计方法

A Channel Estimation Method Based on the DRLNet for OFDM Systems

摘要:

针对OFDM (orthogonal frequency division multiplexing)系统接收信号过程中存在的符号间干扰和子载波间串扰的问题提出了一种高性能的基于深度残差学习网络(deep residual learning network, DRLNet)的信道估计方法。该方法首先在接收端通过最小二乘(least squares, LS)法初步估计导频位置的信道信息，并将该信息作为有噪声的低分辨率数据输入信道估计模型，该模型学习了从导频处有噪信息到完整信道去噪信息的映射关系，从而在模型输出还原的完整信道数据，得到准确的信道状态信息。仿真结果表明: 所提出的DRLNet模型对还原信道状态信息的准确性比传统估计方法更具有优势，在多种信道环境下依然能准确重建信道信息。

Abstract:

A high-performance channel estimation method based on deep residual learning network (DRLNet) is proposed to address the issues of the inter-symbol interference and inter-carrier interference in the signal reception process of OFDM systems. The method first uses the least square(LS) method to preliminarily estimate the channel information at the pilot locations at the receiver, and this information is treated as a noisy low-resolution data to input into the channel estimation model. This model learns the mapping relation from the noisy information at the pilot locations to the complete denoised channel information, thereby the complete channel data at the model is restored and output, thus the accurate channel state information is obtained. Simulation results show that the proposed DRLNet model outperforms the traditional estimation methods in the accuracy to restore the channel state information, and it can still accurately reconstruct the channel information under various channel environments.