Artistic Expressions and Aesthetic Characteristics of Combining Dance and Folk Music

Starting from the current situation of the application of folk music in the art of dance, this paper proposes to utilize artificial intelligence to evaluate the expression and aesthetic characteristics of dance combined with the art of folk music. DeepLabv3+ algorithm is designed for dance movement segmentation, and LSTM model is utilized to process sequence data and complete the recognition of dance movements. The DTW matching algorithm based on cosine distance is proposed to evaluate the dance movements and build the dance evaluation model. Visualize the dance segmentation results using waveform graphs to explore the accuracy of DeepLabv3+ algorithm. Experiment on NTU-RGB+D60 and NTU-RGB+D120 datasets respectively to analyze the dance movement recognition performance of this paper’s model. The accuracy of the evaluation based on the Cosine-DTW algorithm evaluation is verified by comparing the distance between the evaluation output value and the true value. With the help of the evaluator’s satisfaction feedback results, the evaluation effect of this paper’s model on dance combined with folk music art is further judged. The results show that the segmentation point location of DeepLabv3+ algorithm is regular and accurate, and the Cross-Subject and Cross-View accuracy of the proposed model in this paper is 94.3% and 97.9% respectively on NTU-RGB+D60 dataset, and also reaches the optimal level on NTU-RGB+D120 dataset. The user satisfaction of the evaluations generated by the model of this paper is 4.511±0.229, 4.522±0.277, and 4.499±0.306 on the three levels, which is better than other evaluation models.