基于SAE特征优选和Bagging集成学习的信号分选算法OA

Feature selection and classifier design are regarded as crucial steps in radar emitter signal sorting.A novel signal sorting algorithm based on sparse auto encoder(SAE)feature optimization and Bagging ensemble learning is proposed in this paper.First-ly,16-dimensional features are extracted from the intra-pulse,inter-pulse and time-frequency transform domains of radar pulses to form a feature vector,then a SAE-based feature optimization model is established with the vector as the input,and a 5-dimensional optimal feature subset is automatically selected by the model to eliminate redundant information and improve the computational effi-ciency of subsequent sorting algorithm.Subsequently,the optimal feature subset obtained by SAE is adopted as the input to con-struct a Bagging ensemble learning model,which K-means clustering algorithm,mean-shift clustering algorithm and Gaussian mix-ture model are employed as base learners,and density-based spatial clustering of applications with noise is taken as a meta-learner for sorting identification,with final sorting results generated ultimately.Finally,the sorting performance of the proposed method is validated by simulation experiments.It is demonstrated by the results that the optimal feature subset selected by SAE contributes to a significant improvement in sorting performance compared with the direct adoption of high-dimensional raw features.In addition,higher sorting accuracy,lower false detection rate and lower missed detection rate are achieved by the Bagging ensemble model in comparison with individual single model.