Announcement for Downloading full text filePlease respect the Copyright Act.
All digital full text dissertation and theses from this website are authorized the copyright owners. These copyrighted full-text dissertation and theses can be only used for academic, research and non-commercial purposes. Users of this website can search, read, and print for personal usage. In respect of the Copyright Act of the Republic of China, please do not reproduce, distribute, change, or edit the content of these dissertations and theses without any permission. Please do not create any work based upon a pre-existing work by reproduction, Adaptation, Distribution or other means.
URN etd-0829108-115959 Statistics This thesis had been viewed 2821 times. Download 1776 times. Author Wel-Ling Hsu) Author's Email Address No Public. Department Information Management Year 2007 Semester 2 Degree Master Type of Document Master's Thesis Language zh-TW.Big5 Chinese Page Count 80 Title PAT-Tree Based Approximate Search for Query by Humming in Music Retrieval Keyword Approximate Search Fault Tolerance Query by Humming Music Retrieval PAT-Tree PAT-Tree Music Retrieval Query by Humming Fault Tolerance Approximate Search Abstract Query by humming in music retrieval lets user who is unfamiliar with lyric can be free to hum melody as query phrases and retrieves the most similar songs. Because general users are not well-trained professional singers and may not completely keep melody in mind, user could hum query phrases involved in deletion, insertion or transposition errors that increase the difficulties in retrieval.
PAT-Tree is a kind of index structure applied in text retrieval. It converts a sentence of string into a binary string, and then insert all semi-infinite strings in tree. The full text can be efficiently updated and extractd from PAT-Tree. Chien in 1997 [B2] raises the significance of keyword extraction using PAT-Tree based approach, which is efficient in automatic keyword extraction from a set of relevant Chinese documents.
The music database and query phrases used in this research have been preprocessed by tool[A5] provided feature extraction. The phrases are converted to musical note string, on which the retrieval is performed. This research applies PAT-Tree to constructing indexes of musical note string and presents an approximate search with similarity measurement to solve the query errors by humming. The experimental results show that when the error rate of queries less than 40%, the top 5 retrieval accuracy can achieve higher than 90%. It is thus clear that the presented approach has high fault-tolerance.
Advisor Committee Yen-Ju Yang - advisor
Files Date of Defense 2008-07-30 Date of Submission 2008-08-29