In the context of building a strong educational nation, training excellent teachers is an indispensable and important task. The training of teaching skills has become a top priority, and teaching postures directly impact the effectiveness of teaching and the learning experience of students. However, targeted training for these postures has been neglected. Addressing this situation, this chapter proposes a study of the teaching postures of normal university students based on OpenPose. By creating a custom dataset and using ANN (Artificial Neural Network), the teaching postures of student teachers are accurately analyzed, including aspects such as posture and motion fluidity. Compared to traditional training methods, this approach can effectively improve training efficiency and provide a reference solution for universities to address the issues of insufficient teaching staff and excessive teaching pressure in teacher training. Experimental results show that the accuracy of posture detection by our model reaches 96.47%, and the precision reaches 78.86%. Through an in-depth exploration and analysis of teaching postures, more personalized and effective teaching guidance can be provided to student teachers, which is of practical significance for improving their teaching skills.