Research On Multi-style Chord Music Generation Algorithm Based On Artificial Neural Network

With the continuous development of deep learning and artificial neural networks,many related fields have revealed their development potential to people.Among them,algorithmic composition has gradually become a hot research field.Algorithmic composition can not only stimulate the creative inspiration of composers,but also allow a wide range of non-professionals to participate in music creation and enjoy the fun of creation.Tracing back to the development process of algorithmic composition,it is find that with the continuous development of artificial neural network technology,algorithmic composition has also entered a new stage.At present,breakthroughs have been made in many areas of algorithmic composition at home and abroad.Through in-depth research on the algorithms and artificial neural networks proposed in the past,a new algorithm model is proposed to perform algorithmic composition for the generation of chord music,and at the same time process the style of the generated music.The main work and innovations are as follows:（1）Based on the long-short term memory neural network（LSTM）,this paper proposes a grouping-combining algorithm（GCA）model for algorithmic composition,which aims to solve the chord problem in algorithmic composition.The model uses a two-layer LSTM structure,where the first layer of LSTM is used to generate notes,called the note layer,and the second layer of LSTM combines the output of the first layer to generate musical chord accompaniment,called the chord layer.The main purpose of designing this model is to solve the problem of high chord repetitiveness in generated chord music.Through experiments,it is found that the music generated by the model proposed in this paper reduces the chord repetition rate by 3%.（2）After reducing the chord repetition rate of the generated music,it was found that the ideal style of music could not be generated through experiments.In other words,the previous model could not specify the generation of specific styles of chord music.In order to solve this problem,A multi-style chord music generation network（MSCMG）is proposed for creation.This model is improved on the basis of the original model,and added two new modules—music style extraction module and style classifier.The music style extraction module divides the entire music content into two parts,the music style information M_style and the music content information M_content.The style extractor will delete the entangled music style information in the music content information.Through experiments,it is found that the model proposed in this paper can generate musical works in the expected style.（3）A music creation system is designed,and the system has functions such as playing music and music generation.The system is based on the MSCMG algorithm to realize the function of the music generation part,and at the same time allows users to select the type and style of music they want to generate on the page according to their own preferences.The system will select the corresponding music model for music generation according to the different choices of the user,so as to satisfy the user to generate music of the corresponding type and style.The results show that the system can initially meet the needs of users.