Automatic Generation Of Family Music Album Based On Multi-modal Fusion

With the rapid development of electronic technology and big data,the automatic generation of music album is becoming the basic application of people's daily life.In recent years,massive online music and online images have sprang up on the Internet,which greatly improves the user's choice space.How to enable users to enjoy music at the same time with the same emotional photo album pictures has also become the key problem.In order to generate music album easily and enhance user experience of music albums,this paper proposes an automatic generation method of family music album based on multi-modal fusion.This paper first extracts the features of music and images based on the two dimensional emotional model of Thayer and the emotional classification standard of MIREX.On one hand,different from the previous music feature representation method based on file level,this paper uses the musical phrase,the representation mechanism characteristic of sentence level.More emotional information is extracted from the music data,more fine-grained depicts the emotional attributes of music.On the other hand,in order to achieve more effective fusion of two different modal data of music features and image features,this paper proposes a multi-modal LPP algorithm with more emotional segmentation.It maps the music and image features to the implicit space to express multi-modal fusion.In the experiment,the satisfaction and precision rate are used to measure the experimental results,and the results are compared with other methods.At the same time,the results are evaluated on the basis of sentence level feature representation and multi mode implicit spatial representation respectively.Finally,the experimental results show that the proposed sentence level feature representation and the LPP algorithm have achieved good results in the experiment.