| As the “last mile” of LBS(Location-based Services),indoor localization technology has attracted growing number of researchers to get involved.Smartphones,equipped with built-in sensors and supporting a variety of radio frequency signals,can be the best carrier for indoor localization research and application.Until now,the mainstream research scheme of indoor localization,which based on the smartphone,is led by the fingerprint localization base on the RSS(Received Signal Strength)of radiofrequent signal,and auxiliary with multisensory fusion system.After comparing the pro and cons of the current mainstream indoor localization technology,to address this issue that the RSS can only reveal the single-dimension radio frequency signal feature and cannot to accurately perceive the orientation of indoor personnel,this dissertation will adopt the four-directional RSS fingerprint instead of the traditional unidirectional fingerprint,and integrate the behavior perception based on multi-sensor to solve that.In summary,the major result are as follows:(1)To address the problem that traditional location fingerprint matching algorithm can only represent the characteristics of single dimensional fingerprint points,an indoor locating method based on smartphone's Received Signal Strength fingerprint was proposed.Through three steps in the offline stage,including data collection,feature extraction and AP(Access Point)weight assignment,richer fingerprint Point information is extracted.And in the online stage,the improved KNN(K-Nearest Neighbor)algorithm is used to match the test point with the fingerprint point.In addition,experimental study was carried out on a smart phone with Android 10 operating system and 30 test points were randomly selected.Two conclusions can be drawn from the experiment.First,the four-directional RSS fingerprint was superior to the traditional unidirectional RSS fingerprint.Under the same experimental conditions,the four can reduce the positioning error by up to 13.4%.The second is that using the algorithm,which combining the four-directional RSS fingerprint,the average positioning error is around 1.61 meters and the response time is milliseconds.(2)Motion sensing and orientation sensing via the smartphone's built-in triaxial accelerometer and the magneto resistive sensor,and through the behavior perception rules,superimposed the two to conduct indoor personnel's behavior perception.Through the following five kinds of indoor behavior experiments:(1)standing still;(2)walking straight;(3)turning during in walking;(4)running in a straight line;(5)Turning during running.From the experimental results,we can see that behavioral perception can effectively distinguish the five indoor behaviors.(3)In view of the problems such as wall penetration and mutation caused by the traditional fingerprint positioning system,a fingerprint positioning scheme based on fused behavioral perception was proposed.In addition,the scheme adopts the stable and reliable Signals of Opportunity(campus Wi-Fi)after coarse screening and fine screening to build the four-way fingerprint database.The use of Signals of Opportunity will not change the indoor environment,and at the same time,the portability is greatly improved.The orientation perceived during positioning can dynamically switched,and according to the result of behavior perception,the moving speed is determined and the fingerprint location is corrected.The validity of the proposed localization scheme is verified by experiments,and the experiment result shows that: the average positioning error of the scheme using dynamically changed RSS fingerprint database and the improved KNN algorithm based on fuzzy distance is 1.51m,while the average positioning error of the scheme proposed in this paper is 1.21m. |
PDF Full Text Request