Graph and Genre-Based Music Recommendation

For the users who listen to music offline, they may have hundreds or even thousands of songs on their local storage. However, it is difficult for them to select their favorite songs rapidly. To deal with this problem, many researchers proposed corresponding solutions. However, most of these solutions are not very intuitive. In this paper, we propose a graph-based music recommendation system to solve this problem. This system first extracts mfcc and rythm from each song and composes vectors. Then the similarity between each pair of songs is calculated according to mfcc and rythm features. Third, we make use of graph-based visualization technique and interaction tools to show people's current selections and recommendations.