In order to improve the quality of online English teaching, this paper proposes a virtual teacher system based on neural network model. By analyzing the principle and fundamental model of speech conversion technology, the factors that cause speech reliability to decrease linearly due to speech conversion system are obtained. Moreover, a virtual scene model based on graphics and images is established, and the virtual teacher system is constructed by combining the speech recognition algorithm. Then, this paper discusses the core module of virtual teacher, that is, intelligent reasoning and decision-making module. In addition, this paper tries to make an improvement suitable for virtual teachers’ reasoning decision-making in the intelligent decision-making system based on artificial intelligence expert system, and adopts the method of case-based reasoning to make reasoning decision-making. Finally, this paper verifies the effectiveness of this model with experimental research, which has certain reference significance for the subsequent construction of intelligent English teaching system.