Stochastic context-free grammar (SCFG)has a promising application prospect in the field of mode recog-nition and threat estimation of multi-function radars (MFR).The primary limitation of the existing learning algorithms is their huge computing complexity.A fast learning algorithm for the parameters of MFR grammar is proposed,in which the Cocke-Younger-Kasami(CKY)parsing chart is first pre-computed for each training sequence to delete the rules that are not involved in the signal generation.Finally,the estimation of radar grammar parameters is realized with a modified inside-out-side (IO)algorithm.The computing complexity is theoretically analyzed,moreover,simulation experiments are provided to verify the algorithm efficiency.Compared with the conventional IO and Viterbi-score (VS)algorithms,more than half oper-ation time is reduced with our proposed algorithm while the favorable estimation accuracy is maintained.%随机上下文无关文法(SCFG)在多功能雷达(MFR)状态识别和威胁估计中具有良好的应用前景。为了减少常规算法的运算复杂度,本文提出一种基于解析表构造的多功能雷达参数快速估计方法。该方法通过对截获的每个雷达数据序列构造库克-杨-卡塞米(CKY)解析表,排除了大量未参与序列派生过程的产生式,随后在解析表的基础上采用改进的Inside-Outside(IO)算法对雷达文法产生式概率和多功能雷达状态进行快速估计。理论分析与实验仿真证明,该算法在参数估计精度相同的条件下,其运算时间相对于常规IO算法和Viterbi-Score(VS)算法减少了50%以上。
展开▼