臺灣博碩士論文加值系統

抹香鯨(Sperm Whale)為海洋哺乳動物，擁有高度的智慧，會藉由發出不同的聲音來進行互相溝通以及社交行為。欲了解抹香鯨的生態，可以藉由收集鯨豚的聲音資料，進行聲紋的特徵分析與分類。然而在收集鯨豚聲音的過程中，背景雜音也會同時被收錄，因此會降低原始鯨豚聲音的品質，造成聲紋特徵分析與分類的困擾。
本論文針對抹香鯨的點擊(Clicks)、尾波(Codas)兩種聲音類型進行聲紋的特徵分析與分類。聲音來源取自於CIBRA (Centro Interdisciplinare di Bioacustica e Ricerche Ambientali)及TCOML (The Cornel Lab of Ornithology Macaulay Library)的抹香鯨聲音檔。我們對於抹香鯨聲音樣本做背景雜音去除，建立時頻圖，分析聲紋的脈衝間距並且取出最大值、最小值、平均值三個特徵值後，使用模糊C-means分群法做聲紋的分類。實驗結果發現，兩種聲紋都是由不同間隔的脈衝序列所組合而成的，點擊的脈衝間距遠比尾波來的大，脈衝的型態也有顯著的差異。此外，使用模糊C-means分群法可以完全正確的分類。

Sperm whale is one species of odontoceti mammals, possesses highly developed intelligence, performs mutual communication and social activities by sending audio signals of different pulse sequence combinations. To understand the ecology of the sperm whale, we can perform voiceprint’s feature analysis and clustering from the recorded audio signals. However, background noise is corrupted into the original whale’s sound during the sound recording process, thus reducing the quality of the original whale’s sound and leading to the hardship of voiceprint’s feature analysis and clustering.
In this thesis, we focus on two voiceprints, Clicks and Codas, of sperm whale to perform their feature analysis and classification. The source audio files are retrieved from CIBRA (Centor Interdisciplinare di Bioacusticae Ricerche Ambientali) and TCOML (The Cornel Lab of Ornithology Macaulay Library). From these recorded sound files, background noise is suppressed, spectrograms of these two voiceprints are formed, the maximum, minimum and average values of inter-click intervals (ICIs) are extracted as features, and classification is made by fuzzy c-means clustering method. Experimental results show that both Clicks and Codas consist of pulse sequences with different ICIs； Clicks has much larger ICIs than Codas, and their patterns are also quite different. Besides, clustering analysis shows that correct classification can be achieved for all the voice samples.

摘要 I
Abstract II
目錄 III
圖目錄 IV
表目錄 V
第一章 緒論 1
1.1 前言 1
1.2 研究動機與目的 1
1.3 論文架構 1
第二章 基本知識和技術 3
2.1 抹香鯨聲紋介紹 3
2.2 頻譜刪減法 6
2.3 時頻圖(spectrogram) 7
2.4 模糊C-means分群演算法 8
第三章 抹香鯨聲紋之分析 12
3.1 背景雜音去除 13
3.2 時頻分析 15
3.3 選取特徵值建立群聚 17
第四章 實驗結果與分析 19
4.1 Clicks分析結果 19
4.2 Codas分析結果 23
4.3 模糊C-means分群結果 29
第五章 結論與未來展望 35
5.1 結論 35
5.2 未來展望 35
參考文獻 36

[1] Sperm Whale, https://en.wikipedia.org/wiki/Sperm_whale.
[2] CIBRA, http://www-3.unipv.it/cibra/edu_spermwhale_uk.html.
[3] Macaulay Library, http://macaulaylibrary.org/.
[4] http://www.mathworks.com/matlabcentral/fileexchange/27912-vuvuzela-sound-denoising-algorithm.
[5] S. F. Boll, “Suppression of Acoustic Noise in Speech Using Spectral.Subtraction,” IEEE Trans. Acoust Speech , Signal Prcoessing, vol. Assp-27,pp.113-120,1979.
[6] 洪賢昇,“DSP Lecture 3:Transform of LTI System,” 海洋大學電機工程系DSP上課 講義pp.2-10。
[7] Podder, Prajoy, et al. “Comparative Performance Analysis of Hamming, Hanning and Blackman Window, ” International Journal of Computer Applications 96.18, 2014.
[8] James C. Bezdek, Robert Ehrlich, and William Full, “ FCM: The Fuzzy C-means Clustering Algorithm. ” Computers & Geosciences pp.191-203,10.2 ,1984.
[9] http://www.wikiwand.com/zh-tw/取樣定理。
[10] 黃文俊, “模糊權重分群演算法”, 國立台中教育大學教育測驗統計研究所，碩士論文,2009。
[11] 王信文, “抹香鯨聲音時頻特徵分析,” 國立臺灣海洋大學電機工程學系研究所，碩士論文，2015。