基于CSI非视距识别的被动式指纹室内定位

Passive indoor positioning based on CSI NLOS recognition

由于基于WiFi的室内定位技术的主要挑战是多路径和非视距(NLOS),因此,室内定位可以从NLOS识别中获益。然而,来自商用WiFi的接收信号强度信息(RSSI),使得非视距识别受限于有限的带宽和粗略的多径分辨率。文章提出了一种更加细粒度的方法,即利用物理层的信道状态信息(CSI)进行室内非视距识别。利用网卡及现有的WiFi设备,采集室内环境的CSI信息并提取特征信息,构建视距和非视距CSI指纹。利用一种基于深度学习的神经网络算法进行非视距识别,从而达到室内定位的目的。在不同室内环境下的实验结果表明,本方案的NLOS识别率达到96.43%,能有效并准确地区分视距与非视距位置。

Since the main challenge of WiFi-based indoor positioning technology is multi-path and Non Line of Sight(NLOS),indoor positioning can benefit from NLOS identification.However,Received Signal Strength Information(RSSI)from commercial WiFi makes NLOS recognition limited by limited bandwidth and coarse multipath resolution.This paper proposes a more fine-grained method of using the Channel State Information(CSI)of the physical layer for indoor NLOS recognition.Using the network card and the existing WiFi device,the CSI information of the indoor environment is collected and the feature information is extracted,and the CSI fingerprints of LOS and NLOS are constructed.A NLOS recognition based on a deep learning-based neural network algorithm is used to achieve indoor positioning.The experimental results in different indoor environments show that the NLOS recognition rate of this scheme reaches 96.43%,which can effectively and accurately distinguish between the LOS and the NLOS.