mkaasees
09-08-2014, 04:29 PM
Smartphone-Based Collaborative andAutonomousRadio Fingerprinting
[attachment=66825]
Location credibility estimation:
To achieve high credibility of the global fingerprints, the quality of each local RSS fingerprint should be estimated. The location error would increase with the step count due to the error in step length or step detection mechanism. Moreover, as the smart phone is not strapped to a certain position of human body, the sensors typically generate noise when the user generates intense movement. The credibility of each location is estimated by accumulating the standard deviation of acceleration. In the proposed system, we first initialize the credibility at the starting location and then reduce it with the standard deviation of acceleration
Collecting rss:
Mobile users track their position with inertial sensors and measure RSS from the surrounding APs immediately. The position error as well as of the RSS values are stored at the corresponding position in the fingerprint map, which is managed by equally spaced grids. The grid spacing significantly affects the performance of the entire system. A large grid spacing leads to low localization accuracy, whereas a small spacing results in a shortage of time to collect a sufficient amount of RSS samples. In our work, we empirically selected the best grid spacing based on the typical walking speed of mobile users.
Localization of mobile fingerprint:
The mobile user performs self-localization by inertial sensors. Due to the drift error of inertial sensors, the error in the estimated position increases as time advances, and eventually the estimated position will not be reliable anymore. In our system, however, the position of the mobile user is revised with the recursive usage of the RSS fingerprints downloaded from the server. The location of the mobile user is adjusted to the location of the closest matching fingerprint in CF with the currently observed RSS set SOb, and the credibility is also updated as that of the fingerprint. The closest matching fingerprint is the tuple which has the smallest diff.
Tracking a wifi mobile user
We compared the performance of the proposed approach with two traditional approaches; probabilistic approach with 20 RSS samples per grid and the RADAR approach with 1–2 samples per grid. The probabilistic approach estimates the current location using the mean and variation of observed RSS value. A mobile user was walking along the corridor of the building for 200 m. We estimated the location of user and compared the results with ground truth to compute the location error for each approach.
Future Enhancement:
When the user is not serviced with the required information, the user’s details are logged and registered. Later, when the user enters the same location, he is identified by his registered details. The service is provided efficiently as required by the user. Activating prioritization, so that it is possible to provide priorities for selected users among the complex user behavior. Huge number of users utilize the mobile services every day. Some users access specific services frequently. Such users are prioritized over other mobile users.
[attachment=66825]
Location credibility estimation:
To achieve high credibility of the global fingerprints, the quality of each local RSS fingerprint should be estimated. The location error would increase with the step count due to the error in step length or step detection mechanism. Moreover, as the smart phone is not strapped to a certain position of human body, the sensors typically generate noise when the user generates intense movement. The credibility of each location is estimated by accumulating the standard deviation of acceleration. In the proposed system, we first initialize the credibility at the starting location and then reduce it with the standard deviation of acceleration
Collecting rss:
Mobile users track their position with inertial sensors and measure RSS from the surrounding APs immediately. The position error as well as of the RSS values are stored at the corresponding position in the fingerprint map, which is managed by equally spaced grids. The grid spacing significantly affects the performance of the entire system. A large grid spacing leads to low localization accuracy, whereas a small spacing results in a shortage of time to collect a sufficient amount of RSS samples. In our work, we empirically selected the best grid spacing based on the typical walking speed of mobile users.
Localization of mobile fingerprint:
The mobile user performs self-localization by inertial sensors. Due to the drift error of inertial sensors, the error in the estimated position increases as time advances, and eventually the estimated position will not be reliable anymore. In our system, however, the position of the mobile user is revised with the recursive usage of the RSS fingerprints downloaded from the server. The location of the mobile user is adjusted to the location of the closest matching fingerprint in CF with the currently observed RSS set SOb, and the credibility is also updated as that of the fingerprint. The closest matching fingerprint is the tuple which has the smallest diff.
Tracking a wifi mobile user
We compared the performance of the proposed approach with two traditional approaches; probabilistic approach with 20 RSS samples per grid and the RADAR approach with 1–2 samples per grid. The probabilistic approach estimates the current location using the mean and variation of observed RSS value. A mobile user was walking along the corridor of the building for 200 m. We estimated the location of user and compared the results with ground truth to compute the location error for each approach.
Future Enhancement:
When the user is not serviced with the required information, the user’s details are logged and registered. Later, when the user enters the same location, he is identified by his registered details. The service is provided efficiently as required by the user. Activating prioritization, so that it is possible to provide priorities for selected users among the complex user behavior. Huge number of users utilize the mobile services every day. Some users access specific services frequently. Such users are prioritized over other mobile users.