Efficient music genre retrieval based on peer interest clustering in P2P networks

Content-based music retrieval is desirable in Peer-to-Peer (P2P) networks, considering its popularity for users and its ability of semantic search, intensive computing cost raises a barrier to efficiency and scalability though. In this paper, we propose an approach of music genre retrieval based on peer interest clustering. Automatic music feature extraction and adaptive shared music file clustering are described. Peers with similar music genre are clustered together, based on which search mechanism and improvement alternatives are deployed. The results of experiments prove the algorithm increases search performance, including precision and recall while reducing network traffic and peer workload.