Content-Based Music Retrieval in P2P Network

Music files constitute a large portion of the data exchanged in Peer-to-Peer (P2P) networks. However, existing music search systems are limited to title-based search methods, which require us to know exact metadata information such as filenames or artists. In this paper we present a content-based music search system which supports music search only by melody in P2P network. For this we first select a super-peer based P2P network with a hierarchical structure. Then, based on this framework, we efficiently cluster and index all the melodies or peers in the P2P network by indicating the motif representing the music's melodies extracted from MIDI (Musical Instrument Digital Interface) files, as one point (.) in a twodimensional space. Finally, we test our system by producing a small P2P server and client, and simulate a large scale P2P network. The proposed system shows efficiency and scalability even with a large scale network.