Currently, English teaching in colleges and universities faces the challenge of low efficiency in vocabulary teaching. Deep learning technology, an advanced artificial intelligence technology, provides new ideas for solving this problem. Constructing an efficient teaching aid system can achieve personalized vocabulary learning and significantly improve learning efficiency and effectiveness. Aiming to address the efficiency and effectiveness problems in English vocabulary teaching in colleges and universities, this study develops an English vocabulary teaching assistance system based on deep learning. Deep learning algorithms allow the system to analyze students’ learning habits and abilities and provide personalized vocabulary learning suggestions. The research constructs neural network models, uses big data to explore students’ learning behaviors, and develops an intelligent pushing mechanism. The system in this paper achieved face recognition accuracy of 81.08% and 98.84% on the FER and C.K. datasets, which verify the model’s effectiveness. Through the teaching experiments conducted on two classes in Y school in W city, the average score of the English vocabulary test of the class taught with this system increased from 75 to 93, significantly higher than that of the traditional teaching methods. This auxiliary system can improve the effectiveness and efficiency of English vocabulary learning, which also offers new technical support for English teaching in colleges and universities.