快速搜索AFMR特征的雷达信号分选

RADAR EMITTER SIGNAL SORTING BASED ON FAST AFMR FEATURE SEARCH

模糊函数可以对信号结构信息进行较为完整的描述,找出不同信号之间的差异,但搜索信号主脊切面的计算量较大。提出一种混沌优化与差分进化算法相结合的搜索模糊函数主脊AFMR(Ambiguity Function Main Ridge)切面,先差分进化全局搜索,再混沌局部搜索,有效避免了算法陷入局部最优。将提取的信号主脊切面对比穷举法提取的主脊切面,保证正确性的同时显著提高搜索速度。将提取的主脊切面特征进行近邻传播聚类分析,针对聚类算法中偏向参数的不确定,提出动态调整偏向参数的构建式,提高算法性能。实验结果表明,该算法改进的近邻传播聚类准确率在低信噪比的情况下能达到90%以上,明显高于传统近邻传播算法。

Ambiguity function can describe the information of signal structure completely and identify＂ differences between the signals. But it requires huge computational complexity in searching the main ridge slice of signal. In this paper we propose an algorithm combining the differential evolutionary with chaos optimisation for searching the slice of ambiguity function main ridge （ AFMR）, first the global search with differential evolution algorithm, then the local search using chaos, thus effectively avoids falling into local optimum. We contrast signal＇ s main ridge slice extracted by the proposed algorithm with the main ridge slice extracted by exhaustive algorithm, this one ensures the correctness while significantly improves search speed. Affinity propagation clustering analysing is made on the feature of the extracted main ridge slice, we propose the constructive means of dynamic bias parameter adjustment to improve algorithm performance. Experimental results show that the accuracy of the improved affinity propagation clustering can reach 90% and above at low SNR, it is clearly higher than the traditional affinity propagation algorithm.