Musical Element Generation Based On Recurrent Neural Network

Since ancient times,music is one of humanity's hobbies and artistic pursuits.Humans have conducted extensive and in-depth research on music a long time ago,and put forward many music theories.In the history of music,composers did not stop summing up and summing up the techniques of composition.With the development of computer technology and the concept of artificial intelligence,people have created the idea of letting computers perform automatic music creation.Therefore,the concept of algorithmic composition has emerged.The research topic of algorithm composition not only has practicability in the field of music creation,but also allows human beings to sum up their own cognitive and creative processes and has positive significance.Based on the method and thought of machine learning,this article is based on the field of algorithmic composing and aims to study the three essential elements of music:melody,harmony,rhythm,and how to use machine learning methods based on the recurrent neural network to generate these musical elements.The beginning of this article introduces the research background of algorithm composition and introduces the basic knowledge of three kinds of music elements.Then,the authors studied and discussed the methods of generating three kinds of musical elements:First,using the recurrent neural network model to continue the melody writing and the generation of a new melody,the generated melody presents a more coherent structure,and the actual sense of hearing is basically satisfactory;Second,using long-term and short-term memory unit of the circulatory neural network model to generate chords,and compared with the two hidden Markov models,it is concluded that the circulatory neural network model is better than the accuracy of the subjective evaluation of the user The conclusions of Hidden Markov Models;Third,using gradient-frequency neural networks combined with circulatory neural networks for beat tracking,rhythm prediction,and generation,which can achieve high results with accuracy,recall,and F-index measurements.At the end of this paper,based on the background of the rapid development of artificial intelligence,this paper summarizes and looks forward to the method of algorithmic composing.This article is a preliminary study of computer in the field of music creation,which helps to improve the quality and efficiency of music automation creation.It has certain application value in the field of composing,computer-aided composing and pattern recognition.