Music Recommendation Method Based On Audio Characteristics

At present,enjoy online music has become a very common lifestyle.However,the huge amount of music stored on the Internet makes it difficult for people to find their favorite music,which has caused some inconvenience to users.Music recommendation can effectively solve such problems.The music system can efficiently push music that may have interests to users based on analyzing the user's appreciation of music habits,which not only makes the user get better service,but also improves a music.The system's service capabilities and service levels will also be helpful.In this paper,the problem of music recommendation is systematically studied,and a music recommendation method based on audio features is proposed.This method can analyze the similarity of music by comparing the correlation degree of spectrum distribution between music,and it is expected that music with high similarity will be recommended to users.The main work of this paper is music feature extraction and recommendation algorithm design.In terms of music feature extraction,this paper defines Timber,Rhythm and Chroma as the main music features on the basis of consulting a large number of literature and music spectrum distribution.In order to improve the efficiency of music feature extraction and analysis,this paper also optimizes the defined features appropriately,expecting to eliminate some redundant feature components in order to improve the analysis efficiency.As for the recommendation algorithm,music recommendation will become an impossible task if music similarity is calculated only through audio features,because music storage is huge.To this end,we use a hybrid recommendation design strategy.In this paper,a hybrid recommendation algorithm based on FP-growth is presented.The main idea is to analyze users' appreciation habits according to users' history of music appreciation,and then make similarity matching.The design idea is to classify the music in the music library by labels,and then select a certain number of music repertoires with the same label classification.Finally,these candidate tracks are matched by similarity,and a certain number of tracks are selected according to the size of similarity,andfinally recommended to users.The results of this study may provide some support for the development of Internet music recommendation system and help to improve the efficiency of the use of Internet music resources.