Abstract: To address the limited accuracy and high complexity of channel estimation in Orthogonal Frequency Division Multiplexing (OFDM) systems, a Deep Neural Network (DNN) based channel estimation algorithm for OFDM systems is proposed. First, the algorithm replaces the channel estimation and equalization processes relied upon in traditional communication systems with a DNN model. Using pilot information and data blocks at the receiver as input features, the algorithm rapidly learns and estimates channel characteristics. Second, during the model training phase, OFDM frames containing pilot blocks and data blocks are constructed and combined with a wireless channel propagation model to generate a large number of simulated data pairs, ensuring that the training data covers a variety of complex channel environments and noise conditions. Finally, the OFDM system is used to perform a nonlinear mapping of the input-output relationship, transforming it into a task suitable for DNN supervised learning. This enables the trained model to efficiently output accurate channel estimation results at the receiver, reliably recovering the original transmitted data and improving overall system performance. Furthermore, a key computational module for DNN model training is designed based on a field-programmable gate array (FPGA) hardware platform to monitor and optimize the trained model in real time. Experimental results show that compared with the other three proposed algorithms, the proposed algorithm reduces the bit error rate by about 10 times in the medium and high signal-to-noise ratio regions, and the normalized mean square error index is also better than the proposed algorithm. This shows that the proposed algorithm has good anti-interference performance and estimation accuracy under complex channel conditions. In addition, the proposed algorithm consumes moderate resources on the FPGA hardware platform, and the operating power consumption is within a controllable range, verifying the feasibility of deploying the algorithm on FPGA.

Key words: Keywords: Deep Neural Network(DNN), channel estimation, Orthogonal Frequency Division Multiplexing(OFDM), wireless channel, Field Programmable Gate Array(FPGA))

摘要： 针对正交频分复用（Orthogonal Frequency Division Multiplexing, OFDM）系统中信道估计精度受限及复杂度较高的问题，提出一种基于深度神经网络（Deep Neural Network, DNN）的OFDM系统信道估计算法。首先，该算法利用DNN模型替代了传统通信系统中依赖的信道估计与均衡过程，以接收端的导频信息和数据块作为输入特征，从而实现对信道特性的快速学习与估计。其次，在模型训练阶段，通过构建包含导频块和数据块的OFDM帧，并结合无线信道传播模型生成大规模的仿真数据对，以保证训练数据覆盖多种复杂信道环境和噪声条件。最后，利用OFDM系统完成对输入输出关系的非线性映射，将其转化为适合DNN监督学习的任务，从而使训练完成的模型能够在接收端高效输出精确的信道估计结果，实现对原始传输数据的可靠恢复并提升系统的整体性能。同时，设计出基于现场可编程门阵列（Field Programmable Gate Array, FPGA）硬件平台的DNN模型训练的主要计算模块，对训练模型进行实时的监控与优化。实验结果表明，与所提其他三种算法相比，本文算法在中高信噪比区域的误码率指标降低约10倍，归一化均方误差指标也优于所提算法，可见本文算法在复杂信道条件下具有良好的抗干扰性能和估计精度。此外，本文算法在FPGA硬件平台上的资源消耗适中，运行功耗处于可控范围内，验证了该算法部署在FPGA上的可行性。

关键词: 关键词: 深度神经网络, 信道估计, 正交频分复用, 无线信道, 现场可编程门阵列