变学习速率在线ICA算法在雷达信号分选中的应用

Application of Variable Learning Rate On-line ICA Algorithm in Radar Signal Classification

已有的独立分量分析(ICA)雷达分选方法多采用FastICA算法,FastICA算法是一种离线批处理ICA算法,缺乏实时分选能力。文中将一种变学习速率的在线ICA算法应用到雷达分选中,克服了FastICA算法无法实现在线实时分选的缺点;同时,算法能根据相依性测度所反映的信号分离的状态自适应地调节学习速率,平衡了传统在线ICA算法收敛速度和稳态误差之间的矛盾,从而使得采用这种ICA算法的雷达信号分选方法具有收敛速度快,分离效果好的特点。仿真实验验证了分选方法的有效性。

Recently,FastICA is used mostly in radar signal classification.FastICA is an off-line batch processing ICA algorithm and lacks function of real-time processing.In this paper,a variable learning rate on-line ICA algorithm is used for radar signal classification,and this classification method can overcome the disadvantage that FastICA can＇t perform real-time classification.Meanwhile,this ICA algorithm can adjust the learning rate on basis of the degree of signal separation which is reflected by measure of signal dependence,and balance the contradiction between convergence rate and steady error of traditional on-line ICA algorithms.Therefore,the method of radar signal classification using this ICA algorithm has the advantages of fast convergence and better separation performance.Computer simulation results showed effectiveness of the proposed classification method.