Link distance-support vector regression (LD-SVR) based device free localization technique in indoor environment

Abstract

Indoor localization using device free localization (DFL) in wireless sensor networks is gaining momentum nowadays due to the potential benefits of DFL. The techniques used in DFL can be broadly classified as statistical methods, compressive sensing, machine learning, radio tomographic method etc. Whenever loss factor and noise involved in the setup is unpredictable, techniques based on machine learning for target detection improves the result to a greater extend. The adaptability nature of support vector machines eased our choice of machine learning algorithm and support vector machine regression (SVR) is the proposed machine learning approach to address prediction of target position. Proposed link distance-support vector machine (LD-SVR) model uses link distance based DFL of single and multiple targets in indoor environment. Performance of the proposed model using SVR is analysed using parameters mean error and probability distribution function of mean error for various number of nodes and targets by imparting measurement error. The simulation results are found to be very much promising in a 3D room environment. The maximum value of mean error due to measurement error effect on link distance is less than 1 m.

Keywords

• Device free localization
• Machine learning
• RSSI
• Support vector machine regression
• Link distance