Non-orthogonal multiple access (NOMA) is the most emerging radio access scheme for fifth-generation (5G) cellular networks. However, the major drawbacks of present NOMA systems are limited channel feedback and the difficulty of combining them with other advanced adaptive modulation and coding systems. Thus, this paper proposes an optimized NOMA system using a stacked Bi-GRU-based deep learning method for estimating the channel. Initially, the messages are coded by utilizing hybrid polar low-density parity-check (LDPC) code for error-free and efficient coding. The coded messages are then adjusted for frame error rate reduction and data sub-block security using a rate optimization approach, and they are modulated with quadrature amplitude modulation (QAM). Finally, channel estimation is performed using a novel stacked bidirectional Gated recurrent unit (Bi-GRU) technique to manage long-term relationships in information. After transmitting the messages via the estimated channel, decoding is performed by a selective extended segment successive cancellation list (SES-SCL) method on the receiver side. For experimentation, the MATLAB platform is preferred, and the results are evaluated and compared with other existing methods. The comparison analysis demonstrates that the proposed method performs better in terms of outage probability, loss, root mean square error (RMSE), bit rate error (BER), and frame error rate (FER). The obtained findings demonstrate that the proposed mechanism is extremely useful for the NOMA system in providing effective services.

非正交多址 (NOMA) 是第五代 (5G) 蜂窝网络最新兴的无线接入方案。然而，现有 NOMA 系统的主要缺点是信道反馈有限以及难以与其他先进的自适应调制和编码系统相结合。因此，本文提出了一种优化的 NOMA 系统，使用基于堆叠 Bi-GRU 的深度学习方法来估计信道。最初，利用混合极性低密度奇偶校验 (LDPC) 码对消息进行编码，以实现无差错且高效的编码。然后使用速率优化方法调整编码消息以降低帧错误率和数据子块安全性，并使用正交幅度调制 (QAM) 进行调制。最后，使用新颖的堆叠双向门控循环单元 (Bi-GRU) 技术执行信道估计，以管理信息中的长期关系。在经由估计的信道发送消息之后，在接收器侧通过选择性扩展分段连续消除列表(SES-SCL)方法来执行解码。对于实验，首选MATLAB平台，并对结果进行评估并与其他现有方法进行比较。对比分析表明，该方法在中断概率、损耗、均方根误差（RMSE）、比特率误差（BER）和误帧率（FER）方面表现更好。获得的结果表明，所提出的机制对于 NOMA 系统提供有效的服务非常有用。