发明名称 METHOD FOR CHARACTERIZING AT LEAST ONE SIGNAL SOURCE
摘要 A method for characterizing at least one signal source is disclosed. In one aspect, the method includes measuring, from a set of sensors, a set of technical data specific to the signals emitted by the signal source, grouping together the technical data in a set of classes depending on the correlations on the data and characterizing the signal source from technical data having been grouped together. The grouping step includes calculating correlations on the data n-tuples with n>2 and calculating a global partitioning error for different distributions of the data in classes. The global partitioning error is defined as a sum of the partitioning costs for all the n-tuples. The grouping further includes retaining as a grouping of data, the distribution in the classes minimizing the global partitioning error.
申请公布号 US2015074497(A1) 申请公布日期 2015.03.12
申请号 US201414480369 申请日期 2014.09.08
申请人 THALES 发明人 HEURGUIER Dominique
分类号 H04L1/00 主分类号 H04L1/00
代理机构 代理人
主权项 1. A method for characterizing at least one wireless signal source, comprising: measuring, from a set of sensors having antennas, a set of technical data specific to the signals emitted by the signal source; grouping together the technical data in a set of classes depending on correlations on the data; and characterizing the signal source from the technical data having been grouped together; wherein the grouping step comprises: calculating correlations on data n-tuples with n>2;calculating a global partitioning error for different distributions of the data in the classes, the global partitioning error being defined as a sum, for all the n-tuples: of the partitioning costs of an n-tuple in the same class if all the data of the n-tuple are in the same class, the partitioning cost in the same class being a decreasing function of the correlation of the n-tuple; andof the partitioning costs of an n-tuple in different classes when all the data of the n-tuple are not in the same class, the partitioning cost in the different classes being an increasing function of the correlation of the n-tuple; andretaining as a grouping of data the distribution in the classes minimizing the global partitioning error.
地址 Neuilly Sur Seine FR