Segmenting music library for generation of playlist using machine learning

People like listening music primarily due to the emotion it evokes. Any activity or work that a person performs also generates emotions. Considering the above two statements it can assume that people tend to associate music with certain activity if it induces emotions that are in sync with it. In today´s world of infinite storage, the number of songs that a user has is ever increasing. With the increased number of songs, listening music in a way one likes has become difficult. Playlist provide us a way to better organize our music library. These can be created manually or by using various smart playlist creators that generates them by analyzing various components of music files used by the user. In the current scenario the playlist created has a specific number of tracks which are played over and over again, with the only variation occurring being in the order in which songs are played. Also the songs that are not part of any playlist are left redundant, occupying useful memory. Our work focuses on the reducing the number of redundant songs in the music library while creating playlist. We have worked on grouping songs with similar emotional effect together in a segment, and then creating a dynamic playlist every time the user plays music. The results show that as the system is provided with required input, we get a robust playlist for each segment, which consist of diverse mix of songs having similar emotional perception. The playlist created is a non monotonous collection of music which is emotionally supportive to the listener towards the work he is doing.