DSpace university logo mark
Advanced Search
Japanese | English 

NAOSITE : Nagasaki University's Academic Output SITE > Faculty of Engineering > Articles in academic journal >


File Description SizeFormat
DBSJLett_6_4_33.pdf260.42 kBAdobe PDFView/Open

Title: 肺音分類のためのスペクトル分離とロバストな類似度判定による特徴量抽出
Other Titles: Feature Extraction by Spectral Unmixing and Robust Similarity Detection for Lung Sound Classification
Authors: 正田, 備也 / 喜安, 千弥 / 宮原, 末治
Authors (alternative): Masada, Tomonari / Kiyasu, Senya / Miyahara, Sueharu
Issue Date: Mar-2008
Publisher: 日本データベース学会
Citation: DBSJ letters, 6(4), pp.33-36; 2008
Abstract: In this paper, we propose a method for extracting effective features from various lung sounds. We use those features to compose templates useful for lung sound classification. First, we obtain power spectra as feature vectors by FFT. Second, we make feature vector groups partially overlapping with each other and represent each group by a few component vectors provided by spectral unmixing. We put component vectors obtained from various lung sounds into a single set and conduct clustering repeatedly. We can regard the sets of component vectors belonging to the same cluster in all clustering results as effective features. In the experiment, we use a CD accompanying a textbook for nurses and compare clustering results with actual lung sound categories.
Description: 本論文では,肺音分類に有用なテンプレートを作成するために,大量の肺音データから良質な特徴量をふるい分ける手法を提案する.まず,FFT によりパワー・スペクトルを特徴ベクトルとして得る.次に,これらを部分的に重なる複数のグループにまとめてスペクトル分離を適用,各グループを少数の成分ベクトルの組で代表させる.こうして様々な各肺音から得られた成分ベクトルを集め,クラスタリングを多数回実行し,常に同じクラスタに属する成分ベクトル群を良質な特徴量として得る.実験では,看護師用教材CD を使い,クラスタリング結果と実際の肺音のカテゴリとを比較する.
URI: http://hdl.handle.net/10069/16312
ISSN: 13478915
Type: Journal Article
Text Version: publisher
Appears in Collections:Articles in academic journal

Citable URI : http://hdl.handle.net/10069/16312

All items in NAOSITE are protected by copyright, with all rights reserved.


Valid XHTML 1.0! Copyright © 2006-2015 Nagasaki University Library - Feedback Powerd by DSpace