An Enhanced Geometric Filter Algorithm With Channel Diversity for Device-Free Localization

Device-free localization (DFL) techniques estimate a person's location without requiring the person to carry a radio device or tag. In the received signal strength (RSS)-based DFL system, the changes caused by the person's presence in the RSS of links in a wireless network are used to infer his location. A DFL system's accuracy is degraded in cluttered environments due to multipath fading. In this paper, the absolute changes in the RSS of links from different frequency channels are combined into a single measurement vector to mitigate the multipath fading effects. The measurements are then processed by a multichannel geometric filter (MCGF) algorithm that uses link-specific thresholds based on a link's fade level and RSS variance on different frequency channels to detect the target-affected links. In addition, the MCGF algorithm weights probable target locations by the overall fade levels of the target-affected links. Through experiments in both open and cluttered environments, the use of channel diversity and link-specific thresholds is demonstrated to outperform single-channel GF and multichannel methods that are based on radio tomographic imaging in terms of tracking accuracy. The proposed algorithm also has a lower computational overhead compared with existing DFL methods, resulting in much faster execution times.