CSI-Based Probabilistic Indoor Position Determination: An Entropy Solution

Abstract

Location Fingerprinting (LF) is a promising localization technique that enables enormous commercial and industrial Location-Based Services (LBS). Existing approaches either appeal to the simple Received Signal Strength (RSS), which suffers from dramatic performance degradation due to sophisticated environmental dynamics, or rely on the fine-grained physical layer Channel State Information (CSI), whose intricate structure leads to an increased computational complexity. In this paper, we adopt Autoregressive (AR) modeling based entropy of CSI amplitude as location fingerprint, which shares the structural simplicity of RSS while exploiting the most location-specific statistical channel information. On this basis, we design EntLoc, a CSI-based probabilistic indoor localization system using commercial off-the-shelf Wi-Fi devices. EntLoc is deployed in an office building covering over 200 m 2 . Extensive indoor scenario experiments corroborate that our proposed system yields superior localization accuracy over previous approaches even with only one signal transmitter.

Machine Learning for Practical Localization System Using Multiview CSI

“…The systems can discriminate the test CSI data to predict the location labels. For more efficient learning, obtained raw CSIs are applied in advance to various kinds of preprocessing or calibration such as inverse fast Fourier transform (IFFT) [10], linear sanitization [11], [12], kernel density estimation (KDE) [13] or principal component analysis (PCA) [14].…”

AoA-aware Probabilistic Indoor Location Fingerprinting using Channel State Information

“…It outperformed FIFS system with 20% accuracy improvement. Recently, authors in [23] proposed EntLoc, an AR entropy based indoor location fingerprinting system using CSI amplitude information. Compared with PinLoc, the EntLoc's structural simplicity and the strong location-dependency of AR entropy based CSI fingerprint remarkably enhanced the location fingerprinting performance with an average precision improvement of 27.3%.…”

“…If the cumulative contribution rate of the first T taps, i.e., C T , is greater than a predefined threshold C, we then apply a simple rectangular window with length T to truncate the other (K − T ) taps. At last, FFT is further employed on the filtered CIR to yield a smoothed version of CFR [23].…”

“…We thereby adopt the exponentially embedded family (EEF) technique [38] to determine the AR model order due to its superior performance through our extensive experiments. Subsequently, both the AR coefficients and its corresponding model prediction error can be estimated by solving the wellknown Yule-Walker equations using Levinson-Durbin recursion [23]. Once the AR PDF is determined from ( 9), the entropy computation can be then expressed by the following form:…”

“…Despite its structural simplicity, this novel entropy fingerprint embodies the whole statistical information of CSI amplitudes. Through extensive experiments conducted in realistic testbeds, we demonstrate that our proposed AR entropy metric outperforms its original CSI or RSS fingerprint [23]. However, since CSI phases of one subcarrier are generally uniformly distributed [24], this quantifies each RP location with an equally maximized entropy value (a.k.a.…”