A Study on the Lower Bounding to the Dynamic Time Warping-Exampled by Voice Recognition

CHUR > College of Management > Department of Applied Statistics > Seminar Papers > Item 987654321/29522

Please use this identifier to cite or link to this item: http://chur.chu.edu.tw/handle/987654321/29522

Title:	A Study on the Lower Bounding to the Dynamic Time Warping-Exampled by Voice Recognition
Authors:	羅琪 Lo, Chi
Contributors:	應用統計學系 Applied Statistics
Keywords:	動態時軸校正;語音辨識;路徑限制;下限制;時間數列;分類分析;最近鄰居分類法則 dynamic time warping;speech recognition;path constraint;lower bound;time series;classification analysis;nearest neighbor classification rule
Date:	2011
Issue Date:	2014-06-27 00:31:30 (UTC+8)
Abstract:	近年時間數列的分類與群集分析，早已經是資料庫/資料採礦社群中極有興趣研究的主題之一。但大多數的分類與群集方法中都需要計算兩不同時間數列的不相似性量數，但當兩時間數列變動的形狀相似但時間軸並不相同時，歐氏距離等量數對時間軸的扭曲是非常敏感的。使用歐氏距離等量數在分類與群集分析中，雖然計算容易，但不可否認的是會面臨錯誤率較高的現象。而時間軸偏差的問題可以用動態時軸校正(DTW)法解決，此法是一套根基於「動態規劃」的方法，可以有效地校正時間軸的偏差，本研究將採用Fix and Hodges(1951)所提出的 In recent years, classification and clustering of time series data have become one of the topic of great interest within the database/data mining community. Most method in classification and clustering are needed to calculate the distance measure between
Appears in Collections:	[Department of Applied Statistics] Seminar Papers

Files in This Item:

File	Description	Size	Format
s_e313_0082.pdf		27Kb	Adobe PDF	170	View/Open

Loading...