Unsupervised Learning of Finite Mixture Models
IEEE Transactions on Pattern Analysis and Machine Intelligence
Computer
Data Fusion of Power and Time Measurements for Mobile Terminal Location
IEEE Transactions on Mobile Computing
Multivariate Analysis for Probabilistic WLAN Location Determination Systems
MOBIQUITOUS '05 Proceedings of the The Second Annual International Conference on Mobile and Ubiquitous Systems: Networking and Services
Overview of radiolocation in CDMA cellular systems
IEEE Communications Magazine
Hi-index | 0.00 |
Location information is being proposed to assist network management and to provide new wireless services. To accurately locate mobile terminals in dense urban environments and indoors, survey data is used to approximate the joint probability density function (PDF) of the mobile locations and measurements that is used to estimate the mobile location. This method provides accurate mobile terminal location estimates but the cost of collecting the required survey points is high. This paper introduces the use of the Variational Bayes technique to approximate the joint PDF of locations and measurements as a Gaussian Mixture Model with a prior distribution of model coefficients. The parameters of the prior distribution are common for multiple survey sets that are collected in network areas with known commonalty such as from the same building or city neighborhood. Methods are presented for calculating these parameters by using data from the survey sets so that the Mean Square Error (MSE) of mobile location error is minimized. The use of prior distributions creates information sharing when calculating the parameters of the joint PDFs for each survey set. This reduces the number of survey points in each survey set required to achieve a desired accuracy. The efficiency and robustness of this method to different measurement noise distributions is demonstrated for a simulated dense urban environment.