Dance education has been continuously promoted from standardized movement training to emotional experience, artistic expression and intelligent evaluation. Focusing on the optimization of dance teaching driven by emotional resonance, this paper constructs an intelligent analysis model that integrates multimodal data preprocessing and synchronous alignment, posture recognition, emotional feature coding, resonance gating fusion and sequential attention aggregation, and designs teaching intervention and intelligent feedback mechanisms. The verification was carried out on the basis of 64 students and 8 weeks of teaching experiment. The results showed that the comprehensive expression of the experimental group increased from 70.26 to 86.57, and the emotional resonance increased by 19.39. The Accuracy of the model reaches 91.42%, the F1-score reaches 90.73%, and the average response time of the system is controlled at 2.8 s. The research shows that this method can improve the accuracy of emotion recognition, performance evaluation and feedback guidance in dance teaching, and provide an operable and interpretable technical path for the cultivation of artistic expression.
