Uyghur dance has a long history and needs to rely on education to inherit. However, the Uyghur dance education in local colleges and universities does not follow the development of the times. Aiming at the situation of Uyghur dance education in colleges and universities, a Uyghur dance pose generation network is constructed. By analyzing the main components and tasks of the autoencoder, the process of extracting music and dance movement features is explored. Combine LSTM to retain the information of dance action features and solve the problem of disappearing dance gradients. Extract the dance video as a sequence of skeletal key points and replace each joint point position with a number accordingly to improve the representation of dance action. The Uyghur dance pose generation network is used to analyze the features of Uyghur dance and propose promotion strategies based on the dance features. The results show that the amplitude and power of female Uyghur dance movements are relatively small, and the distance between joints and points is only [10,90] cm, while the amplitude of male Uyghur dance movements is much larger, and the distance between joints of the point rubbing step is [-10,100] cm. in the promotion of the dance, the degree of validity of the promotion of the combination of modern technology and Uyghur dance is 0.8546. The present study has promoted the reform of Uyghur dance education to some extent. The reform of Uyghur dance education is conducive to the preservation of national traditional dance.