In order to solve the problem of emotion absence in multimedia English teaching, an intelligent network teaching system model based on facial expression recognition is proposed. Principal component analysis (PCA) is applied to extract the important feature frames of facial expressions in online learner videos; the facial emotion recognition network based on CNN architecture judges and understands the emotional state of the learner, and gives corresponding emotional encouragement or emotional compensation strategies according to the specific emotional state of the learner. Simulation results show that the average detection rate of the proposed algorithm is 78.28%, and the average recognition accuracy is 81.78%, compared with VGG16 and ResNet50.