Signal parameterization vs. orthogonalization on example of vehicle's magnetic signature recognition [road traffic measurement]

Often the task of road traffic parameter measurement systems is the recognition of the class of vehicles moving on the road. The quality of such recognition depends on many factors: kind of sensor used, quality of measurement channel, recognition method used to analyze generated signals corresponding to vehicles. In the presented paper, an inductive loop sensor and vehicle magnetic signature signal are described. The paper presents two approaches to recognition. The first one consists in replacing the original signal with a vector of its parameters and looking for the minimum distance between parameter vectors and pattern vectors in each class of vehicle. A second approach relies on the search for minimum distance between the original tested signal and the first component of K-L (Karhunen-Loeve) expansion in every vehicle class. Results of both methods of recognition for the same set of signals are compared by a common, author defined, factor.