Abstract: Intelligent decision-making is the core of anti-jamming communication system，the optimal decision is made on the system's jamming suppression mode, spectrum resource allocation, modulation and coding mode and power adjustment information according to the jamming environment. The existing intelligent decision-making in the anti-jamming communication system mostly adopts the genetic algorithm, artificial bee colony algorithm, etc. In the face of complex and changing electromagnetic environment, usually these algorithms do not have the fault tolerance capability for environment estimation parameters and the generalization ability for the new jamming. The BP neural network algorithm is simple, has certain fault tolerance and generalization ability. In this paper, a real-time anti-jamming decision engine model based on BP neural network is designed and analyzed. The pre-processing method and discriminant standard of input data are designed according to system performance. The decision-making steps and the algorithm parameters are analyzed. The system performance simulation proves that the decision engine proposed in this paper has strong anti-jamming performance. Compared with the decision engine using genetic algorithm or artificial bee colony algorithm, the decision engine proposed in this paper is faster and has generalization ability and fault tolerance.