English  |  正體中文  |  简体中文  |  Items with full text/Total items : 28588/40619
Visitors : 4185973      Online Users : 47
RC Version 4.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Adv. Search
LoginUploadHelpAboutAdminister

Please use this identifier to cite or link to this item: http://ntour.ntou.edu.tw:8080/ir/handle/987654321/29981

Title: 多尺度生物聲學之複雜度研究:從個體到群落
The Complexity of Multiscale Bioacoustics: Organism to Community
Authors: Yu-Hsiang Pan
潘宇翔
Contributors: NTOU:Department of Environmental Biology and Fisheries Science
國立臺灣海洋大學:環境生物與漁業科學學系
Keywords: 生物聲學;MSE多尺度熵;SKD滑動樹;高品質聲音;ICA獨立元分析
bioacoustics;high quality sound;independent component analysis (ICA);multi-scale entropy (MSE);sliding kd tree (SKD)
Date: 2011
Issue Date: 2011-11-25T03:32:55Z
Abstract: 生物聲學(Bioacoustics)因其可靠、非侵入性的特點,近年在生物個體、族群、群聚甚至生物多樣性的監測上都引起大量的關注。但它在以時間為主軸的訊號分析使用上,卻必須面對不同層面的挑戰。如個體之生物之心電圖(ECG)、腦電圖(EEG)、眼電圖(EOG)、長週期溫度變化、或生態之長週期1/f變化,一般統計方法已不敷使用,研究人員雖可以利用多尺度熵(Multisacle entropy, MSE)方法研究複雜的長時間序列數據分析找出相關性,但卻必須面臨龐大計算量所導致之記憶體不足、速度不夠快的難題;從族群與群聚之生態監控角度而言,生態學家必須設法從原來嘈雜的聲音訊號中,得到良好目標物種聲音結構,並同時留存當時的棲地之背景音景(Soundsacpe) 為了解決個別生物體、族群及群聚生態之研究限制,本研究針對發展出新的演算法,將傳統視窗搜尋(Window search)的方法,轉為計算幾何方法(Orthogonal range search),並進一步自創滑動KD樹(Sliding kd tree, SKD)及自適應SKD樹(Adaptive SKD)演算法,以解決複雜度計算的龐大數據計算量問題。執行時間從使用線性記憶體的前提下,從數據長度平方計算量次O(N2),降為O(N1-1/d),當d=2時,為O(N3/2)。使用在數位整數之生物醫學數據時,數據同計算量次O(N),已可達到線上即時監控的效果,並成功運用在生醫領域之睡眠診斷上,未調教前即可達到76%的準確度;在機械震動診斷及VLSI佈局輸出上,也有相當突破。 群聚之生態研究方面,則建議採取獨立元分析(Independent Component Analysis, ICA),先回顧了聲音訊號濾波處理,實驗室模擬實驗、實驗室預錄但野外實播之半野外實驗,及純野生動物青蛙野外聲音鳴叫之實測,成功記錄目標動物的完整聲音,並保留當時之棲地背景音。 此生物聲學複雜度之ICA聲音分離技術與MSE快速演算的改善,可順利應用在個體生物之觀察研究,結合ICA後,更可有效記錄、觀察群聚生態活動,使人類更容易發現更多生物、生態及環境的秘密,同時也對生物資源及環境資源的可持續發展,做出有意義的貢獻。
Bioacoustics has drawn a great deal of attentions for monitoring system health situation of organism, population, community, and species diversity with it’s reliable, non-invasive characteristics. But it also has different challenges and complexities organism’s very large and long-term time series data of biology, such as signal of electrocardiogram (ECG), electroencephalography (EEG), interbeat interval (RR), electro-oculography (EOG), long-term temperature changes data, or 1/f noise of ecology such as population dynamics etc. Although a researcher can use approximate entropy (AE), sample entropy (SE) or multi-scale entropy (MSE) to examine the complexity of the huge data, unfortunately, the researcher must first solve those high difficulties, expensive execution time in computation of the MSE or AE. In automatic monitoring species diversity, an ecologist must takes the challenges to reduce the sound complexity, getting the fine structure sound and keep the background sound of habitat from the original sound with noise recorded in wildlife. To solve the limitations of bioacoustics, the author suggests a new algorithm called sliding kd tree (SKD) to fast up the computational time of MSE. The author in this thesis first try to compute multiscale entropy with orthogonal range search of kd tree (KD), to develop a new algorithm by SKD, and successful to reduce the computation time cost of MSE from O(N2) to O(N3/2) when d=2. Moreover, if used in digital typed data such as biomedical data, the execution time can advance be improved to O(N), which is fast enough for on line diagnosis. Besides, he also successful applied MSE in biomedical signal of sleeping EEG online diagnosis and achieved to the correct rate of 76% or higher before any tune, and other areas such as machine vibration and optimal two dimensional orthogonal range search in VLSI automated design. To solve the noise problem of ecology monitoring, the author suggests independent component analysis (ICA) to separate the focal animal and back ground habitat sound. He first reviewed the acoustic signal filter processing, experimenting with wildlife frog’s sound, suggests filtering to enhance the loudest or louder sound, and suppress the other sounds, which is based on ICA method and using two microphone to record the focal animal’s sound and then use ICA to separate the focal sample sound and the other noises for high quality sample sound. With the new algorithm of MSE, SE and AE and new ICA filter in wildlife, bioacoustics can then get more chance to be implemented in more areas of organism to community. This may lead humanity easier to discover more biology and ecology secret, and be helpful to manage the biological resources and the sustainable development of nature resources.
URI: http://ethesys.lib.ntou.edu.tw/cdrfb3/record/#G0D95310001
http://ntour.ntou.edu.tw/handle/987654321/29981
Appears in Collections:[環境生物與漁業科學學系] 博碩士論文

Files in This Item:

File Description SizeFormat
index.html0KbHTML214View/Open


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

 


著作權政策宣告: 本網站之內容為國立臺灣海洋大學所收錄之機構典藏,無償提供學術研究與公眾教育等公益性使用,請合理使用本網站之內容,以尊重著作權人之權益。
網站維護: 海大圖資處 圖書系統組
DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback