Radar source identification method based on sample reduction and improved support vector machine

摘要

Aiming at the problem of low efficiency of radar emitter identification method, a new method based on sample reduction and improved support vector machine is studied. Firstly, for removed redundant information, at the same time reduce the training data, the algorithm through the local normal vector to boundary extraction of sample prior information in the database. Then using the Sequential Minimal Optimization algorithm, multi classification and cross-validation to improve the original SVM. Through the improved algorithm train the reduced samples, and get the optimal model parameters. Finally using the optimal identification model to recognize the unknown pulse sequence information. Through simulation results and comparison, it is proved that the proposed radar source identification method based on sample reduction and improved support vector machine not only have high identification accuracy and robustness, but also have a good timeliness.