Baek et al. proposed a deep learning-based signal detector for conventional MIMO systems, which is a pioneering work of applying artificial intelligence into wireless communications. Although this work works well under static fading channels, it is worth notable that it fails to work under block-fading channels. In particular, the detection BER under block-fading channels is around 0.5 even in the high regime of SNR. To explain this unexpected result, we provide some simple yet efficient theoretical analysis, which clearly verifies that the proposed detector cannot decouple the phases between the channel parameters and transmitted signals and hence it fails to detect the transmitted signals under block-fading channels. The results in this paper can help understand and improve the detector. In particular, the detector structure should incorporate the channel state information for the application under block-fading channels.

中文翻译：

---

关于传统 MIMO 发射机基于深度学习的信号检测的实现方法的说明

贝克等人。为传统的MIMO系统提出了一种基于深度学习的信号检测器，这是将人工智能应用于无线通信的开创性工作。虽然这项工作在静态衰落信道下工作良好，但值得注意的是它在块衰落信道下无法工作。特别是，即使在高 SNR 状态下，块衰落信道下的检测 BER 也约为 0.5。为了解释这个意想不到的结果，我们提供了一些简单而有效的理论分析，它清楚地证明了所提出的检测器不能解耦信道参数和传输信号之间的相位，因此它无法检测块衰落信道下的传输信号。本文的结果可以帮助理解和改进检测器。特别是，