In view of traditional data-centric radar warning technology failure in building up a model of Multi-function Radar at dynamic state and analyzing signals effectively, a mathematical model is established by utilizing stochastic context free grammar(SCFG) to signal production mechanism of MFR, and on this basis SCFG production rules and MFR′s state transition matrix are estimated by EM algorithm, MFR′s operation state is estimated by Viterbi algorithm. The paper also studies the effect of grammar initial values on the algorithm. The simulation results show that the use of the method can effectively estimate MFR system parameters under condition of giving a proper initial value, and the correct probability of the state estimation can come to about 92%.