Novel Pattern Indexing Techniques for Efficient Content-basedMusic Retrieval
碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 96 === With the progress of multimedia capturing techniques, multimedia data, especially music data is more and more tightly bound up with our life. How to access the user-interested music efficiently and effectively is a hot topic for content-based music retrieval....
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2008
|
Online Access: | http://ndltd.ncl.edu.tw/handle/59835405576670483363 |
Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 96 === With the progress of multimedia capturing techniques, multimedia data, especially music data is more and more tightly bound up with our life. How to access the user-interested music efficiently and effectively is a hot topic for content-based music retrieval. Because music data is temporal, many previous works used not only the low-level features of music itself but also the sequence of the features. Most of the works put their concentration on enhancing the accuracy but ignore another important issue: the efficiency. In our method, the high dimensionality of low-level features is first reduced to improve the retrieval performance by the labeling approach. Then we make use of the temporal continuity to find the valuable patterns within a sliding window and construct the index structure. Through our proposed method, the music that is related to user’s query can be found rapidly.
|
---|