Device-Free Localization And Tracking In Wireless Sensor Networks

As one of the key techniques in wireless sensor networks(WSNs),device-free local-ization(DFL)and tracking,which can detect and locate a person without the need for any wireless devices,plays a more and more important role in military,medical,and security areas.However,with the development of the DFL and tracking,there are sev-eral challenges:(1)the improvement of the localization and tracking accuracy,and(2)the low computational complexity in device-free tracking.Hence,further research and the accuracy improvement in DFL and tracking are the keys to the target localization and tracking.To investigate the DFL in different environments,we propose two algorithms,i.e.,a new elliptical model for radio tomographic imaging(RTI)algorithm in outdoor envi-ronments,the fingerprint-based DFL in changing environments using enhanced channel selecting and logistic regression(indoor environments).To improve the tracking accu-racy,an RSS-distance-angle weighted geometric filter is proposed.The contributions of this paper are as follows:(1)There is still much room for improvement in DFL in outdoor environments.To address this problem,the RTI algorithm is researched and the orthogonal match-ing pursuit algorithm(OMP)is used to improve the localization accuracy.Firstly,the wireless channels inside the elliptical models,which represents communication links,are divided into line-of-sight and non-line-of-sight paths.The monitoring area are divided into voxels and the elliptical models are divided into several different areas that rep-resent different weights,which is more consistent with the actual situation.Secondly,as for the ill-posed inverse,the OMP algorithm is firstly used in DFL to derive an im-age estimator,and the extra bright spots can be eliminated.The experimental results demonstrate that the proposed algorithm can futher improve the localization accuracy;(2)When the environment changes,e.g.,the bad influence of the wall on the com-munication signals,and the movement of the furniture,there is still much room for localization accuracy improvement in DFL.To fill this gap,the fingerprint algorithm is researched and the logistic regression classifier is utilized to improve the localization accuracy.Firstly,the link with the highest average RSS is selected,and the Pearson correlation coefficients of pairs of the communication frequency channels in that link are computed during training.Secondly,with the same procedures,the Pearson correlation coefficients of the frequency channels are computed during testing.A pair of channels with the highest Pearson correlation coefficient is selected,and the data collected from the two channels will be utilized both in the training and testing procedures,which would be robust to the environmental change.Finally,the logistic regression classifier,which can better process highly dimensional data,is firstly used in device-free localiza-tion to generate the estimated position.The experimental results demonstrate that the localization accuracy has been improved by the proposed algorithm,without the need for rebuilding the training database;(3)As the extension of the DFL,device-free localization and tracking has signifi-cant practical value.However,there is still much room for improvement in localization and tracking accuracy with low computation complexity.To solve this problem,the geometric-based algorithm is researched and an RSS-distance-angle weighted model is proposed to improve the localization and tracking accuracy.Firstly,when the target is in the monitoring area,the change of RSS(?RSS)in each communication link is computed.The communication links that outside the estimated area(the relationship between the estimated positions and the prior estimated position)are filtered by the set of the ?RSS threshold.The ARSS of the rest links that inside the estimated area will be treated as the RSS-based weights of intersection points of the links.Secondly,the distance-based weights that are based on the distance between the prior estimated position and the intersection points are proposed.The smaller the distance between the prior estimated position and the intersection point is,the bigger the value of the distance-based weight should be.Finally,to ensure the correct direction of motion,we propose the angle-based weights,which are based on the angle between the prior estimated position and the intersection point.The direction with the most number of intersection points will be treated as the correct direction of motion.The experimental results demonstrate that the proposed algorithm can improve localization and tracking accuracy.