The application of human-computer interaction (HCI) technology in dance teaching is an emerging direction to improve teaching quality. Traditional dance teaching relies on teachers’ demonstration and verbal instruction, which makes it difficult for learners to get immediate feedback and affects the mastery of complex movements. This paper establishes a dance movement recognition model through a computer vision system, explores the impact of human-computer interaction technology on the learning efficiency of complex dance movements, and provides technical support and innovative methods for dance teaching. The study adopts image preprocessing, thresholding processing and background subtraction techniques to process the dance video, and combines the mean Hausdorff distance, Hidden Markov Model and joint point angle measurement to construct the dance movement recognition system. The experimental part trains the 2D human key point recognition model with 3D coordinate regression model, which is validated on MPII and Human3.6M datasets. The results showed that: the average recognition rate of the constructed dance movement recognition model for ten complex movements reached 98%; the total dance skill scores of students in the experimental class applying the system were significantly higher compared to the control class with traditional teaching (89.78±3.849 vs. 84.215±4.869, P<0.01); and the three indexes of musical rhythmic accuracy, expressiveness, and body form retention all presented significant differences (P<0.01); 80% of the students in the experimental class showed strong or very strong self-confidence, compared with only 40% in the control class. The study proves that dance teaching based on human-computer interaction technology can significantly improve learners' learning efficiency, skill level and learning interest, and provides a new teaching mode and method for dance education.
