Application of Machine Learning in Automatic Generation of Intangible Cultural Heritage Music Scores

In order to achieve an organic combination of inheritance and innovation in the field of intangible cultural heritage music, this paper conducts a comprehensive analysis of the application of machine learning in the automatic generation of intangible cultural heritage music scores. The deep neural network technology model is used to implement the training activities based on the traditional intangible cultural heritage music data set. The task aims to develop an automatic generator to generate music scores that conform to the traditional music style. The original data is preprocessed and feature extraction is completed. The features of pitch, rhythm and harmony in the music elements are extracted and analyzed. The deep neural network model is used for training and performance adjustment optimization scheme, and the quality of the generated image is comprehensively improved. The generated music scores show a high degree of similarity with the original scores in terms of pitch, rhythm and harmony, and the rhythm density and note change frequency show obvious coherence. The created music scores faithfully reflect the structural style of classical music. This expression shows a certain innovative quality. The work has both artistic aesthetics and practical functions.