基于膜粒子群算法的雷达辐射源信号多目标特征选择方法研究

A study of a multi-target feature-selection method of radar signal sources based on membrane particle swarm optimization

针对粒子群算法易陷入局部最优和收敛性能较差问题,受膜计算优化理论启发,提出1种膜框架下的粒子群算法,用于解决无监督的多目标雷达辐射源信号特征选择问题.在表层膜中,采用非支配排序和拥挤距离机制使算法既保留了多目标粒子群优化算法的快速收敛性,同时使用基本膜的进化规则也使解集具备较好的多样性,避免了过早收敛问题.在雷达辐射源信号特征提取中,利用一阶差分自相关方法提取包络特征,采用相关度和冗余度两个目标优化数据对象,以评价雷达辐射源信号特征子集的质量,并应用于雷达辐射源信号的脉内特征选择.实验结果表现出算法具有较好的可聚类性和全局收敛性,特征信号之间明显可分,边界清晰无交叠,获得了更高的雷达源信号分选识别正确率.

Aiming at the problem that particle swarm optimization is easy to fall into local optimal and poor convergence performance,and inspired by the theory of membrane computing optimization,a particle swarm optimization algorithm under membrane framework is proposed to solve the feature-selection problem of unsupervised multi-target radar sources.In the surface layer membrane,the non-dominated sorting and crowded distance mechanism are adopted to keep the fast convergence of the multi-objective particle swarm optimization algorithm,and the evolutionary rules of the basic membrane are used to make the solution set have better diversity,avoiding excessive early convergence problems.In the feature extraction of radar emitter signals,the first-order differential autocorrelation method is used to extract the envelope features,and the two objects of correlation and redundancy are used to optimize the data object to evaluate the quality of the feature subset of the radar emitter signal and apply Intrapulse feature selection of radar radiation source signals.The experimental results show that the algorithm has good clusterability and global convergence,the feature signals are clearly separable,the boundaries are clear and there is no overlap,and a higher recognition accuracy of radar source signal sorting is obtained.