海洋研究常需要將儀器做長時間佈放, 以取得季節性的變化。但是受限於儀器記憶體容量及電池電量的限制, 如果長時間連續作業, 系統在有限的資源下, 則無法連續擷取所需之資料。 因此必需發展一套可針對收集資料做偵測及篩選的分類器,使得系統在佈放期間能針對有用的訊號做收集並減少非實驗所需的水下資料。本研究將針對偵測出訊號中屬於鯨豚的資料, 以供相關研究單位使用。
本論文資料的收集是由一自主式水下聽音器 (Passive Acoustic Listener, 簡稱PAL) 由美國華盛頓大學應用物理實驗室的J. Nystuen 博士所發展的一套水下聽音系統。 PAL為一非連續性紀錄的聽音器, 當環境聲音產生的頻譜能量達到一設定強度時, 則開始前置設定週期為4.5秒的紀錄, 反之, 則處於休眠狀態, 等待下一次週期的開始。 以往研究鯨豚類發聲的方法, 不外乎是針對單純的鯨豚訊號做時頻域的分析,其中包含頻譜分析 (Spectrogram correlation) 及動態時間校正 (Dynamic time warping) 等等。但針對實海域長期收集到的水下訊號而言, 包含許多自然界的聲響, 如風、 雨、 船舶和生物等訊號, 僅靠頻率分析無法明確地找出鯨豚特徵。然而, 在處理頻譜圖時, 卻可藉由影像展現之特徵及聲強值的能量變化有效區分出鯨豚類。 就像是我們可以很容易地從陌生人當中分辨出哪些人是曾經見過一樣,故我們假設在一群未知訊號中並不需要所有資料點的演算即可區分出其中的類別。

為簡化此問題, 我們引進以點為基礎的主成份分析法, 和傳統主成份分析法的演算不同於這個概念建立在取出訊號的特徵,以每個分析框產生的特徵資料來代表原始資料, 再與訓練資料庫中的分析框, 進行主成份分析演算, 最後利用歐幾里得距離 (Euclidean distane) 來進行分類。經測試證明兩種特徵抽取方法的正確率分別可達到78\%及93\%,成功地驗證此改良式的演算法可有效地偵測出鯨豚的存在。

For many undersea research application scenarios, instruments need to be deployed for more than one month which is the basic time interval for many phenomena. With limited power supply and memory, management strategies are crucial for the success of data collection. For acoustic recording of undersea activities, in general,either preprogrammed duty cycle is configured to log partial time series,or spectrogram of signal is derived and stored,to utilize the available memory storage efficiently.To overcome this limitation, we come up with an algorithm to classify different and store only the sound data of interest.

Features like characteristic frequencies, large amplitude of selected frequencies or intensity threshold are used to identify or classify different patterns. On main limitation for this type of approaches is that the algorithm is generally range-dependent, as a result, also sound-level-dependent. This type of algorithms will be less robust to the change of the environment.One the other hand, one interesting observation is that when human beings look at the spectrogram, they will immediately tell the difference between two patterns. Even though no knowledge about the nature of the source, human beings still can discern the tiny dissimilarity and group them accordingly. This suggests that the recognition and classification can be done in spectrogram as a recognition problem. In this work, we propose to modify Principal Component Analysis by generating feature points from moment invariant and sound Level variance, to classify sounds of interest in the ocean. Among all different sound sources in the ocean, we focus on three categories of our interest, i.e., rain, ship and whale and dolphin.

The sound data were recorded with the Passive Acoustic Listener developed by Nystuen, Applied Physics Lab, University of Washington. Among all the data, we manually identify twenty frames for each cases, and use them as the base training set. Feed several unknown clips for classification experiments, we suggest that both point-based feature extraction are effective ways to describe whistle vocalizations and believe that this algorithm would be useful for extracting features from noisy recordings of the callings of a wide variety of species.

1 緒論. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1
1.1 研究動機. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 相關研究. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 研究目的. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 論文架構. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 主成份分析法. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7
2.1 主成份分析概述. . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 主成份分析法的流程. . . . . . . . . . . . . . . . . . . . . . . 9
2.3 改良式主成份分析法-以點為基礎. . . . . . . . . . . .11
3 資料收集及處理. . . . . . . . . . . . . . . . . . . . . . . . . . . .19
3.1 儀器規格. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1.1 取樣參數設定. . . . . . . . . . . . . . . . . . . . . . . . . . .20
3.1.2 佈放環境. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21
3.2 鯨豚的聲音訊號. . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3 聲音訊號前處理. . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3.1 訊號處理流程. . . . . . . . . . . . . . . . . . . . . . . . . . .24
3.3.2 訊號框分析. . . . . . . . . . . . . . . . . . . . . . . . . . . . .25
4 特徵向量. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27
4.1 運算流程. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.2 訓練資料. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 影像不變矩. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3.1 影像前處理. . . . . . . . . . . . . . . . . . . . . . . . . . . . .32
4.3.2 不變矩(Moment Invariant)產生特徵向量. . . .42
4.3.3 影像辨認. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44
4.4 聲強變異數. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.4.1 計算聲強變異數. . . . . . . . . . . . . . . . . . . . . . . . .45
4.4.2 平滑處理(Smoothing) . . . . . . . . . . . . . . . . . . .49
4.4.3 特徵點選取. . . . . . . . . . . . . . . . . . . . . . . . . . . . .51
4.4.4 訊號辨認. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .56
5 測試實驗. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . .57
5.1 實驗規劃. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .57
5.2 評估指標. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . 58
5.3 不變矩特徵測試. . . . . . . . . . . . . . . . . . . . . . . . . . .59
5.4 變異數特徵測試. . . . . . . . . . . . . . . . . . . . . . . . . . .61
6 結語. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.1 討論. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .64
6.2 結論. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . 68
6.3 建議. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .69
附錄. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
A 影像不變矩辨認流程與步驟. . . . . . . . . . . . . . . . . . .73
B 聲強變異數辨認流程與步驟. . . . . . . . . . . . . . . . . . .77

[1] J. A. Nystuen, ”Listening to raindrops from underwater: An acoustic disdrometer,”J. Atmos. and Ocean. Tech., vol. 18, pp. 1640-1657, 2001.
[2] M. E. Dahlheim and F. Awbrey, ”A classification and comparison of vocalizations of captive killer whales (Orcinus orca) ,” J. Acoust. Soc. Am. 72, pp.
661-670, 1982.
[3] J. K. B. Ford , ”Call traditions and vocal dialects of killer whales (Orcinus orca) in British Columbia,” Ph.D. Dissertation (University of British Co,umbia, Vancouver).
[4] J. K. B. Ford , ”Acoustic behaviour of resident killer whales (Orcinus orca)off Vancouver Island, British Columbia,” Can. J. Zool. vol. 67, pp.727-745, 1989.
[5] J. K. B. Ford , ”Vocal traditions among resident killerwhales (Orcinus orca)off Vancouver Island, British Columbia,” Can. J. Zool. vol. 69, pp.1454-1483,1991.
[6] V. B. Deecke, J. K. B. Ford and P. Spong, ”Quantifying complex patterns of bioacoustic variation: Use of a neural network to compare killer whale (Orcinus orca) dialects, ” J. Acoust. Soc. Am. 105,pp. 2499-2507, 1999.
[7] J. C. Brown, A. H. Davis and P. J. O. Miller, ”Classification of vocalizations of killerwhales using dynamic timewarping,” J. Acoust. Soc. Am. 119 (3), 2006.
[8] M. Turk and A. Pentland, ”Eigenfaces for recognition,” Journal of Cognitive Neuroscience, vol. 3, no. 1, pp. 71-86, 1991.
[9] L. A. Cowan andH. FaridNASA SHARP PLUS Summer Program, ”Eigenfaces for Face Recognition,” 1995.
[10] C. C. Wang, H. C. Hsu and M. S. Zeng,” Hand Signal Recognition for Diver-Machine Interface,” in the Oceans of IEEE, vol. 2, pp.1231-1236, 2002.
[11] 許宏昌, ”主成份分析法於指尖亮點手勢辨識之應用”, 國立中山大學海下技術研究所碩士論文, 2003。
[12] K. Pearson,”On lines and planes of closest fit to systems of points in space,”PhilosophicalMagazine, 2, pp. 559-572, 1901.
[13] H. Hotelling, ”Analysis of a complex of statistical variables into principal components,” Journal of Educational Psychology, 24, pp. 417-441, 1933.
[14] G. M. Wenz,”Acoustic ambient noise in the ocean : spectra and sources,” J. Acoust. Soc. Am. 34, pp.1936-1956, 1962.
[15] J. A. Nystuen, ”Temporal Sampling Requirements for Autonomous Rain Gauges”, J. Atmos. and Oceanic Tech.,vol. 15, pp. 1254-1261, 1998.
[16] B. B. Ma, J. A. Nystuen, and R. C. Lien, ”Prediction of underwater soun levels from rain amd wind,” J. Acoust. Soc. Am. 117, pp. 3555-3565, 2005.
[17] F. Thomsen and D. Franck, ”Characteristics of whistles from the acousticrepertoire of resident killer whales (Orcinus orca) off Vancouver IslandBritish
Columbia,” J. Acoust. Soc. Am. 109, pp. 1240-1246, 2001.
[18] M. K. Hu, ”Visual Pattern Recognition byMoment Invariances,” IEEE Trans. Info. Theory, vol. 8, pp. 179-187, 1962.
[19] S. A. Dudani, K. J. Breeding, and R. B. McGhee, ”Aircraft identification by moment invariants,” IEEE Trans. Comput., vol. C-26, no. 1, pp. 39-45, 1983.
[20] J. H. Li, Q. Pan, P. L. Cui, H. C. Zhand and Y. M. Cheng,”Image Recognition Based on Invariant Moment in the Projection Space,” IEEE Proceedings of the Third International Conference on Machine Learning and Cybernetics, Shanghai, 26-29, 2004
[21] N. Otsu, ”A threshold selection method from gray-level histograms,” IEEE Trans. System,Man, and Cybernetics, vol.SMC-9, pp. 62-66, 1979. 71
[22] T. C. Hsia, ”A Note on InvariantMoments in Image Processing,” IEEE Trans. System,Man, and Cybernetics, vol. 11, pp. 831-834, 1981.
[23] 繆紹綱, 數為影像處理活用Matiab , 全華科技圖書股份有限公司, 1999年11月,頁2-10。
[24] J. F. Canny, ”A computation approach to edge detection,” IEEE Trans. On Pattern Analysis andMachine Intelligence, vol. 8, pp. 679-698, 1986.
[25] R. Gonzalez and R. Woods Digital Image Processing, Addison-Wesley Publishing Company, pp. 442, 1992.