Signal Perturbation Based Support Vector Regression for Wi-Fi Positioning
Location estimation using Received Signal Strength (RSS) in pervasively available Wi-Fi infrastructures has been considered as a popular indoor positioning solution. However, accuracy deterioration due to uncertainty of RSS and offline manual calibration cost limit the deployment of Wi-Fi positioning systems. This paper proposes a signal perturbation technique to enhance existing Support Vector Regression (SVR) based Wi-Fi positioning. By signal perturbation, more RSS training samples are generated, thus enhancing the generalization ability of SVR. In addition, Access Point (AP) selection method is applied to reduce the input dimension by discarding the redundant APs.