MRTB framework: a robust content-based music retrieval and browsing

We propose new content-based music retrieval and browsing system called MRTB. New feature extraction procedures such as multi-feature clustering (MFC) and sequential forward selection (SFS) are proposed for robust and optimum performance of the music retrieval system. Using MFC and SFS, we can achieve 86% of average retrieving success rate and 0.5 ∼ 0.8 sec retrieval time per each 20 s query based on top 10 ranked list rules. The user friendly graphical interface of MRTB is implemented in a thumbnail manner so the user can easily verify their retrieving results with some useful musical information.