Korean folk song retrieval using rhythm pattern classification

In content based music retrieval, approximate string matching approaches are frequently used. Though string matching algorithms effectively deal with the input fault tolerance of user query, they require much time to search the entire database. In this paper, we propose a Korean folk song retrieval method which efficiently search the database. A user query is searched in two steps, a coarse classification using rhythm pattern and an interval comparison based on approximate string matching. The experimental results show that the proposed method efficiently improve the retrieval time.