High-accuracy localization technology has gained increasing attention in gesture and motion control and many diverse applications. Due to multi-path fading and blockage effects in indoor propagation, 0.1m-level precise localization is still challenging. Promising for 6G wireless communications, the Terahertz (THz) spectrum provides multi-GHz ultra-broad bandwidth. Applying the THz spectrum to indoor localization, the channel state information (CSI) of THz signals, including angle of arrival (AoA), received power, and delay, has unprecedented resolution that can be explored for positioning. In this paper, a Structured Intra-Attention Bidirectional Recurrent (SIABR) deep learning method is proposed to solve the CSI-based three-dimensional (3D) THz indoor localization problem with significantly improved accuracy. As a two-level structure, the features of individual multi-path rays are first analyzed in the recurrent neural network with the attention mechanism at the lower level. Furthermore, the upper-level residual network (ResNet) of the constructed SIABR network extracts hidden information to output the geometric coordinates. Simulation results demonstrate that the 3D localization accuracy in the metric of mean distance error is within 0.25m. The developed SIABR network has very fast convergence and is robust against THz indoor line-of-sight blockage, multi-path fading, channel sparsity and CSI estimation error.

中文翻译：

---

SIABR：一种用于超精确太赫兹室内定位的结构化注意力内双向循环深度学习方法


高精度定位技术在手势和运动控制以及许多不同的应用中越来越受到关注。由于室内传播的多径衰落和阻塞效应，0.1m级的精确定位仍然具有挑战性。太赫兹 (THz) 频谱提供多 GHz 超宽带宽，有望用于 6G 无线通信。将太赫兹频谱应用于室内定位，太赫兹信号的信道状态信息（CSI），包括到达角（AoA）、接收功率和延迟，具有前所未有的分辨率，可用于定位。本文提出了一种结构化内注意力双向循环（SIABR）深度学习方法来解决基于CSI的三维（3D）太赫兹室内定位问题，并显着提高了精度。作为一个两级结构，首先在循环神经网络中分析各个多路射线的特征，并在较低级别上使用注意机制。此外，所构建的SIABR网络的上层残差网络（ResNet）提取隐藏信息以输出几何坐标。仿真结果表明，以平均距离误差为指标的3D定位精度在0.25m以内。所开发的SIABR网络具有非常快的收敛速度，并且对太赫兹室内视距阻塞、多径衰落、信道稀疏和CSI估计误差具有鲁棒性。