Reduction Of Individual Transmitter Features Based On Neighborhood Modle

Individual Transmitter Identification is a very important research topic in the field of information warfare, it mainly depends on signalprint to determine which transmitter that the signals come from so as to to achieve its locking, tracking, surveillance, electronic jamming and even military strikes purposes. Unlike the traditional direction of signal research, Individual Transmitter Identification mainly focus on slight differences of transmitters of same type even same batch and then to extract and analyze.As an important basis of Individual Transmitter Identification, signalprint is one of the most important part of the whole process. However, the current signal fingerprint has a wide range of different extraction methods, different classification results, so the selection and combinations of properties of fingerprint has become a major problem. This article uses the neighborhood rough set theory, overcoming the problem of traditional rough set theory which can only deal with discrete data, based on the concept of reduction and postive region,has produced the optimal feature subset and experiments with an neighborhood classifier, the result shows that the feature subset has a similar or even better classification ability compared with the original one.The main work of this thesis includes:1) extracted a variety of signalprints, combined the fingerprint with the neighborhood rough set theory, used attribute reduction algorithms based on neighborhood rough set model to process the continuous data which the classical rough set can not handle,providede a new way of finding the combinations of signalprints 2) designed and simulated an experiment to implement the reduction operation based on neighborhood rough set, proved the feasibility and effectiveness of experimental design by using the neighborhood rough set classifier.3) using measured signal data, according to experimental results, compared the multiple reduction algorithms and analyzed the identification results.