臺灣博碩士論文加值系統

摘要
海洋生態礙於天候、環境、地形等因素，常使學者無法順利從事海洋生態觀
測的研究。理想的情況下，透過水下收音系統收集鯨豚的語音資料，便能分析鯨
豚的生態；然而在現實的條件下，海底收音常參雜雨聲、浪聲、風聲及其他生物
的聲音，導致原始收集的鯨豚語音之時頻圖無法明確辨識鯨豚特徵。
為了解決辨識上的困難，本論文以時頻分析與影像處理的方式來完成。首先
對於原始收集的虎鯨語音進行時頻圖的建立，而後，經由影像濾波、特徵區域選
取、二值化等處理，最後求取質心的位置，以有效辨識不同物種在接近頻段的語
音影像之差異。實驗結果顯示，所提方法能有效辨識三種族群transient(過境
型)/Offshore(離岸型)/Resident(常駐型)虎鯨的脈衝語音特徵。

Abstract
Due to the factors of weather, environment, topography, oceanographers are
often unable to observe or research easily on the ocean ecology. Normally it will be
able to analyze whale’s and dolphin’s ecology by using passive acoustic listener
(PAL) to collect their voice data. However, in actual situations, the voice data
collected by PAL usually include the sounds of raindrops, waves, wind and other
living beings, thus causing its original voice spectrogram hard to identify the whale’s
and dolphin’s features clearly.
In this thesis, we use the spectrogram method and some specific image
processing techniques to overcome this identification problem. First, the
spectrogram of the original orca whale’s voice data is constructed. Next, image
filtering, feature region selection, and binarization are performed to extract the
centroid position, so that the image differences of different species at the nearby
frequencies can be identified effectively. Experimental results show the three
ethnic groups, namely Transient/Offshore/Resident, of orca whales can be identified
effectively from their pulsed call voice data.

目次
摘要 I
Abstract II
目次 III
圖次 IV
表次 V
第一章 緒論 1
1.1前言 1
1.2研究動機與目的 3
1.3論文架構 4
第二章 研究背景 5
2.1虎鯨簡介 5
2.2頻譜刪減法 6
2.3離散傅立葉轉換 7
2.4短時距傅立葉轉換 8
2.5漢明窗 9
2.6Matlab影像處理函式 10
第三章 研究方法與實驗結果 11
3.1前言 11
3.2語音與影像處理 12
3.3特徵提取 19
3.4實驗結果 24
第四章 結論與未來展望 26
參考文獻 27
附錄 29

參考文獻

[1] Judith C. Brown and Patrick J. O. Miller, “Automatic classification of killer whale vocalizations using dynamic time warping”, Acoustical Society of America, 2007.
[2] Judith C. Brown and Andrea Hodgins-Davis and Patrick J. O. Miller, ” Classification of vocalizations of killer whales using dynamic time warping” Acoustical Society of America, 2006.
[3] http://seaworld.org/animal-info/animal-infobooks/killer-whale/communication-and-echolocation/
[4] F. Thomsen and D. Franck, J. K. B. Ford, “Characteristics of whistles from the acoustic repertoire of resident killer whales (Orcinus orca) off Vancouver Island, British Columbia,” J. Acoust. Soc. Am. , pp. 1240-1246, 2001
[5] 潘榮貴, “晶圓塗料過程異常音辨識之研究”, 國立臺灣科技大學電機工程系論文, 2007
[6] Saeed V. Vaseghi, Advanced Digital Signal Processing and Noise Reduction, Second Edition., John Wiley &; Sons Ltd, chapter 11, page 334-354
[7] A.V Oppenheim, R.W. Schafer, Discrete-Time Signal Processing, Prentice Hall, Inc.,1989
[8] http://cnx.org/contents/e1744717-b8c5-4e28-80e0-9e0c4ee8389c@22/Spectrograms
[9] http://mirlab.org/jang/books/audiosignalprocessing/speechFeatureMfcc_chinese.asp?title=12-2%20MFCC
[10] W.B. Hussein, M.A. Hussein and T. Becker, “Spectrogram Enhancement By Edge Detection Approach Applied To Bioacoustics Calls Classification,” DOI:10.5121/sipij.2012.3201
[11] http://www.mathworks.com/matlabcentral/profile/authors/1711356-choqueuse-vincent
[12] http://www.mathworks.com/products/matlab/