Classifying the Geometric Dilution of Precision of GPS satellites utilizing Bayesian decision theory

The errors resulting from satellite configuration geometry can be determined by Geometric Dilution of Precision (GDOP). Considering optimal satellite subset selection, lower GDOP value usually causes better accuracy in GPS positioning. However, GDOP computation based on complicated transformation and inversion of measurement matrices is a time consuming procedure. This paper deals with classification of GPS GDOP utilizing Parzen estimation based Bayesian decision theory. The conditional probability of each class is estimated by Parzen algorithm. Then based on Bayesian decision theory, the class with maximum posterior probability is selected. The experiments on measured dataset demonstrate that the proposed algorithm lead, in mean classification improvement, to 4.08% in comparison with Support Vector Machine (SVM) and 9.83% in comparison with K-Nearest Neighbour (KNN) classifier. Extra work on feature extraction has been performed based on Principle Component Analysis (PCA). The results demonstrate that the feature extraction approach has best performance respect to all classifiers.

The paper deals with the implement of the proposed feature extraction and classification approaches for the purpose of GPS GDOP analysis. The Parzen algorithm estimates the conditional probability of each class. Then based on Bayesian decision theory, the class with maximum posterior probability is selected.Figure optionsDownload as PowerPoint slideHighlights
► The new GPS GDOP classification method based on both Parzen estimation and Bayesian decision theory has been proposed.
► The conventional GPS GDOP features have correlations through t-test.
► The novel feature extraction method for analyzing GPS GDOP has been developed utilizing Principle Component Analysis.