A Method for Fast States Estimation of Multi-function Radar Based on Syntactic Derivation of Parse Chart

Abstract

Stochastic Context-Free Grammars (SCFG) have promising application prospect in the field of Multi-Function Radars (MFR) states recognition and threat estimation, which entails the fast learning of the probability of radar grammar production based on training data. Conventional learning algorithms are limited in practical application for their high computational complexity. A new fast learning algorithm for the probability of MFR grammar is proposed in this paper in light of a reformulated and delicate SCGF modeling of the MFR signal generation mechanism. The proposed algorithm first pre-compute Cocke-Younger-Kasami(CKY) parsing chart for each training sequence, and then the probability of radar grammar production is estimated with modified Inside-Outside(IO) algorithm based on the aforementioned parsing chart. The computational complexity and accuracy of the algorithm are also analyzed in detail. Simulation results show that the algorithm could fast estimate production rule probabilities with favorable estimation accuracy, and compared with the conventional IO or Viterbi Score (VS) algorithm, more than a half operation time can be reduced.