| Summary: | 碩士 === 國立臺灣師範大學 === 資訊教育學系 === 95 === With a great deal growth of digital contents such as media data of voices、musics、images and videos,it becomes an important topic for digital contents retrieval.The music retrieval focuses on the part of musics,and users retrieve songs from internet or database by humming、voicing and keying.In this day,people are no longer restricted to get resources in words with the maturation of media techniques. Accordingly,more and more pictures、voices and images would be objects to be searched.
Music retrieval is based on the core of string matching presently.There are not only exact matching but approximate matching(allowing errors).However,it’s would take large number of time when it proceeds with huge datas.Take exact matching,m denotes the number of querying notes and n denotes the number of music notes,and the time complexity is O(m × n).
String matching is a normal and significant problem. We are searching for the set of starting positions F = {i | 1≦i≦n-m+1 ,titi+1...ti+m-1 = P} for a string P=p1p2...pm inside a large text file T=t1t2...tn.The characters may be English characters in a text file, DNA base pairs, lines of source code, angles between edges in polygons, music notes and tempo in a musical score, and so forth. The two most
famous algorithms for this problem are the Boyer-Moore algorithm [13] and the Knuth Morris Pratt algorithm [14].
This thesis adopts Fast Text Searching Allowing Errors[6] as theorem to achieve music retrieval.It’s very suitable to apply on hardware with Shift、AND、OR algorithm only,and improves time cost significantly.m denotes the number of querying notes and n denotes the number of music notes.The hardware time complexity of the algorithm is O(n) irrelevant to the number of allowing errors(d) while O (d × n) for software.We append our circuits on Stratix development board by Altera corp. accompany with user querying UI(user interface) coded by Borland C++ 6.0 composed with network forms a high-speed FPGA music retrieval system.
|
|---|
