BDS Multipath Signal Classification Using Support Vector Machine

In urban environments, multipath can significantly deteriorate the positioning precision of the global navigation satellite system (GNSS). BeiDou navigation satellite system independently established by China plays an important role in the GNSS market. Eliminating the multipath is a crucial problem to contribute to the development of the BeiDou navigation satellite system (BDS). In this paper, we use the machine learning algorithm support vector machine (SVM) to classify the BeiDou satellite signals into line-of-sight (LOS), multipath, and non-line-of-sight signals (NLOS). Single and multiple feature classification of the signal was performed by using the carrier to noise ratio (C/N0), elevation angle (ELE), and pseudorange residuals (PR). We use SVM with radial basis function (RBF), which can effectively handle nonlinear and high-dimensional data, and this feature is just suitable for the effective classification of nonlinear and high-dimensional data in this paper. It is a challenging problem to select the appropriate features from receiver independent exchange (RINEX) format signals for the diverse forms of signals output from BeiDou signal receivers. In this paper, we analyze the selected features C/N0, ELE, and PR, and it is proved that they can be used for BeiDou satellite signal classification. In the experimental study, BeiDou satellite signals are collected with static receivers in an urban canyon. The experimental results show that the highest classification accuracy of 78.48% is achieved based on the PR using a single feature aspect. The SVM classification accuracy based on feature C/N0, ELE, and PR can reach 87.22%. The classification using multiple features is significantly higher than that of single feature.