CSI-Based Signal Reconstruction for WiFi Localization

CSI-Based Signal Reconstruction for WiFi Localization

연구 분야: Infrastructure

학회: International Conference on Internet of Things as a Service

With the advent of the 5G era, the demand for high-precision wireless positioning continues to grow. However, traditional ranging-based positioning systems are highly susceptible to interferences caused by multipath and none-line-of-sight (LOS) propagation, which can significantly degrade the accuracy of the estimated time-of-arrival (TOA) values. To address this challenge, this paper proposes a Deep neural networks (DNN)-based approach for accurate TOA estimation in indoor environments. Using a complex-values neural network model, the proposed method predicts TOA directly from the frequency domain channel state information (CSI) of wideband WiFi receivers. We also propose an input normalization method based on peak search in the channel impulse response, which improves both the accuracy of TOA estimation and the efficiency of model training. The proposed method was verified experimentally both in an outdoor area of 900 \(m^2\) with 6 anchors and an indoor area of 700 \(m^2\). It is shown that the proposed approach significantly outperforms conventional methods, with 77% of the positioning errors within 0.5 m in the outdoor test and 95% within 1 m. In the indoor test, about 64% of the positioning errors were within 0.5 m, and approximately 80% were within 1 m.