基于最优解析树提取的多功能雷达状态快速估计方法

摘要

针对基于文法建模的多功能雷达（ Multi-Function Radar ，MFR）参数估计领域中常规算法具有的高运算复杂度问题，提出一种快速估计算法。该算法利用文法的派生过程仅与文法结构有关，而与文法概率参数无关这一事实，利用库克-杨-卡塞米（Cocke-Younger-Kasami，CYK）算法对截获雷达数据序列进行预处理，构造出可以反映该序列派生过程的解析表，进而从该解析表中提取出序列的最优解析树，然后利用改进的Viterbi-Score算法对雷达文法概率参数进行快速估计。论文仿真分析了该算法的计算复杂度、存储复杂度和估计精度，实验结果表明了该算法相对于常规算法，可以减少60％左右的计算量。%To deal with the huge computing burden of the existing multi-function radar ( MFR) syntactic model pa-rameters learning algorithms,a fast learning algorithm is proposed in light of the derivation only relevant to the syntactic ar-chitecture but the probabilities.In our method, each training sequence is pre-processed by the Cocke-Younger-Kasami ( CYK) parsing algorithm,the parse chart is constructed to accurately describe the sequence derivation.Furthermore,the best parse tree is extracted from the parse chart,and the probabilities are estimated based on the best parse tree with a modified Viterbi-Score algorithm ( VS) .The time complexity,memory complexity and accuracy are also explored.Simulation results show that compared with the conventional algorithm,more than 60%operation time can be reduced with our proposed algo-rithm.