| Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 98 === Traditionally, people retrieve preferred music by computing the similarity of music content, namely content-based music retrieval (CBMR). A number of studies on content-based music retrieval have been presented until the present. However, it is not easy to make a high precise search within a query session. It motivates us to develop a query refinement technique called PBRF (Pattern-based Relevance Feedback) that combines three kinds of query refinement techniques, namely QPM (Query Point Movement), QR (Query Reweighting) and QEX (Query Expansion). To deal with the local optimal problem, we additionally propose a novel switch-based search strategy that adaptively selects the best search strategy based on user’s feedbacks. Through the integration of QPM, QR, QEX and switch-based search strategy, the user’s intention can be captured more precisely in the global search space. The experimental results reveal that our proposed approach performs better than existing CBMR methods in terms of precision.
|
|---|
