隨著人類工業化的加深，不僅對於海洋的污染日益嚴重，長久以來的濫捕濫撈，更對海洋生態造成毀滅性的破壞。為了不讓漁業資源枯竭，諸多計畫管理型漁業的概念順勢興起，其中又以海洋牧場(Ocean Ranch)最受世人矚目。要讓海洋牧場能夠成功運作，藉以復甦水中生機的關鍵在於人工魚礁的設置。其作法是先由岸上水產養殖單位孵育魚類幼苗，再施放到沿海區域，魚苗經由箱網養殖到一定大小之後，讓魚群棲息於人工魚礁，使得海洋牧場的構想不因魚群四散而失敗。當然，要設置人工魚礁，最重要的在於魚礁基礎是否穩固，因此施放地點的海床底質成分就需事先調查清楚。本研究計畫採取間接調查的方式，以主動式聲納—測深儀(Depth Sounder)發射聲波訊號，再分析由海床底質反射回來的訊號(Reflecting Signals)，來瞭解人工魚礁施放地點的海床底質成分。

但是各種聲波訊號在水中的傳播過程中，受到海洋環境干擾與水中各種噪音滲入的影響，訊號本身雜亂不堪，因此在接收後必須接著進行適當的前置處理與頻譜分析，萃取出較具有代表性的特徵參數，進而分類並辨識出海床底質成分。而最後這分類與辨識的工作，本研究計畫是交由類神經網路(Artificial Neural Network, ANN)來執行。類神經網路隨著電腦硬體科技、認知科學、生物科學與社會科學等相關理論的日益成熟，再加上傳統人工智慧(Artificial Intelligence, AI)技術發展遭遇瓶頸而再度獲得重視。類神經網路處理資訊的方式，跟人體大腦中的腦神經架構一樣，具有快速運算能力、高記憶容量、學習能力、容錯能力等特色，並具備歸納推廣、分散式平行處理、自我組織與自動調整的強大功能。面對數量龐大、雜亂無章的水下訊號，類神經網路算是既快速、又有效率的分析處理工具，因此已成為發展新一代人工智慧系統的重要技術。

在眾多網路運算模式中，本研究計畫決定採用目前最受廣泛使用的倒傳遞式(Backpropagation)學習演算法，作為類神經網路的主體架構。目的在於辨識出三種海床底質成分：細砂（或沈泥）(Fine Sand)、砂(Medium Sand)與岩石(Rock)。在聲波訊號的處理過程中，特徵參數的萃取是採用峰值選取的方式，將各種底質的峰值頻率參數導入類神經網路進行訓練與學習，再根據辨識結果與實際結果的偏誤關係，調整網路的權值，以達辨識成功率改進的效果，最後建構出確實可行的即時聲波訊號辨識系統。

Along with advancement of human industrialization, pollution in the ocean is getting worse. Moreover, the overfishing through the years has caused catastrophic damage to the ocean eco-system. In order to avoid exhaustion of fishery resource, many concepts of planned administrative fishery has become popular, and thereamong, ocean ranch draws the most attention. Artificial reef plays a key role in an ocean ranch, which starts with incubating brood fish in the laboratory. Often, the brood fish will grow in the cage near coast till proper size, then be released to the artificial reef. If fish groups do not disperse and multiply, the artificial reef can be considered successful. The success of the artificial reef relies on the stable foundation. Consequently, the composition of seabed sediment under the planned site should be investigated thoroughly before hand. This research introduced a remote investigation method, which an active sonar, depth sounder, was used to emit and collect acoustic signals. By using the signals reflected from the seabed, the sediment composition can be analyzed.

However, all acoustic signals are subjected to noise through propagation, and distorted somehow. Therefore, certain signal pre-processing should be applied to the received signal, and representative characteristics can be extracted from it. In this research, the recognition platform was built on artificial neural network (ANN) in this research.

Among many network algorithm modes, this research chose the widely used backpropagation learning algorithm to be the main structure in ANN. The goal of this research was to discriminate among three seabed sediments: fine sand, medium sand, and rock. During the signal processing, characteristics were extracted by using peak value selection method. Selected major frequency peaks were fed into the network to train and learn. According to partial error relation between recognition and practical result, weights of the network were adjusted for improving successful ratio. Finally, a reliable acoustic wave signal recognition system was constructed.

誌謝…………………….……………………………………………………………...I

摘要…………………….……………………………………………………………..II

Abstract…………………….………………………………...……………………..III

第一章 緒論…………………….……………………………………………….1

1.1 研究背景……………………………………………………..….………....….1
1.1.1 海洋牧場的源起…………………………………………………….1
1.1.2 人工魚礁的重要性………………………………………….………4
1.2 研究動機與目的…………………..……………………………….………….8
1.2.1 直接調查…………………………………………………………….8
1.2.2 間接調查…………………………………………...…………….….9
1.2.3 海下探測儀器介紹……..………….…………..…………….…….10
1.2.4 海床底質調查方式的比較………..…………….…………………11
1.2.5 海床底質聲學性質的研究歷史………………..……..….…..……12
1.2.6 台灣四周海床底質資料的蒐集……………………….….….……13
1.3 研究方法………….……….…………….……………………..……..……...14

第二章 訊號模式識別方法………………………………….………………...15

2.1 數位訊號處理與模式識別…………………...………………….……..……15
2.2 樣本映對的抽象表示法…………...……………..…..………….……..……16
2.3 模式識別系統內部架構…………………….…………………...……..……18
2.4 常用的模式識別方法……………………………………………………..…20
2.5 模式識別方法的比較………………………..….………………………..….23
2.6 類神經模式識別方法…………………………...………………………..….24

第三章 聲波訊號處理……………………………………….……………..….26

3.1 訊號取得…….……..……….…….…...………………..…....………...…….26


3.1.1 海洋三號……..…….………………………………………..……..26
3.1.2 測深儀……………………....………………………………..…….29
3.1.3 數位錄音機…..…….…………………………………….…..…….30
3.1.4 其他工具……..…….………………….……………………..…….31
3.2 訊號前置處理…….……..……….………….....………………….…...…….34
3.3 加窗與正規化處理……………………………….….…….…...…….....…...39
3.4 頻譜分析與特徵擷取…….…….…...……………..……....………...……....41
3.5 比對辨識…….……..……….…….…...……………..……....………...…….43

第四章 類神經網路……..……….………………………..…...………....……44

4.1 類神經網路的緣起…………………..…………………………...……….…45
4.2 類神經網路理論介紹.….……………………...………….…...………....….46
4.3 類神經網路的分類…..………………………………………...………….…49
4.3.1 以網路架構來分類………………………………………………...49
4.3.2 以學習策略來分類………………………………………………...50
4.4 類神經網路的特色…………………..……………………...…………….…53
4.4.1 快速運算能力……………………...……………..………….…….53
4.4.2 高記憶容量………………………………………………….……..55
4.4.3 學習能力…………………………………………………….……..55
4.4.4 容錯能力…………………………………………………….……..55
4.5 類神經網路的應用.…………………………..…..……………………….…57
4.5.1 依問題類型分類……………………….…………………….…….57
4.5.2 依應用領域分類…………………………………………….……..58
4.6 類神經網路程式撰寫……….…….….……………………………...………59
4.7 類神經網路模式選取……………….……………………………….………60
4.8 倒傳遞類神經網路……………….………………………………………….61
4.8.1 發展過程……………………………………………...……………61
4.8.2 網路模型………..……….……….…………………...….…….…..61
4.8.3 優缺點比較與問題討論………………………....……..…..……...64

第五章 底質粒徑分析…………………………………………….…………...68

5.1 沈降管分析實驗……………………………..………………………….…..68
5.1.1 原理………………………………………………………...………68
5.1.2 使用儀器與設備…………………………...………………...….…70


5.1.3 MacRSA程式系統的簡介…………………………………………71
5.1.4 實驗步驟一：樣本前處理…………………………………………72
5.1.5 實驗步驟二：快速沈降分析RSA實驗………………..………....72
5.2 庫爾特粒徑分析實驗……………….………....……………………………75
5.2.1 庫爾特原理……………..…………….……...……………….……75
5.2.2 庫爾特原理儀器特性……………………...………………………76
5.2.3 庫爾特LS系列—雷射粒徑分析儀……………………………….76

第六章 研究結果與比較……………………….……………………………...80

6.1 實驗結果……………………………………………………….………....….80
6.1.1 實海作業…..…….……..…………...…...………………….………80
6.1.2 室內作業……………..……………………….………....…….……83
6.2 類神經網路辨識結果……………………………………………….……….88
6.2.1 細砂的辨識結果……………………………………………………..90
6.2.2 砂的辨識結果…………………………………………………….….91
6.2.3 岩石的辨識結果……………………………………………………..92
6.3 結果討論……..………………………………………..……………...……...93
6.3.1 聲波頻率…..…….……..……………..…….….……...….….………93
6.3.2 分類辨識軟體…..…..……….………………….……...…….………94
6.4 結論與展望………………………………………………………...…..….....95

參考文獻……..………….……...…..…….…………..……………….………..…....96

附錄一：倒傳遞類神經網路系統程式（找出加權值與偏權值）….…...…....……..100

附錄二：倒傳遞類神經網路系統程式（辨識驗證之用）….…...…..……….……...103

[1] 呂台生與高源國, ”海洋牧場之工程技術應用,” 經濟部八十年度科技發展
專案計畫調查報告, 1991.
[2] 張崑雄, 人類與海洋, 教育部高中高職環境教育課外讀物叢書之16, 教育部
環保資訊網路之規畫及推廣計畫, 臺大地理系地圖與多媒體研究室, 1996.
[3] 歐慶賢與翁平勝, ”海洋牧場的相關技術與經營管理(上),” 臺灣省漁業局
漁業推廣第127期, pp. 13-26, 1997.
[4] 張崑雄, “人工魚礁,” 中央研究院動物研究所專刊第一號, 1976.
[5] Mathews H., Physical and Geological Aspects of Artificial Reef Site Selection. In
Artificial Reefs: Marine and Freshwater Application, Lewis Publishing Inc.,
Chelsea, 1985.
[6] 歐錫祺, 鄭火元, 劉仁銘, 陳惠玲與黃基紘, “花蓮縣石梯坪人工魚礁區海
域環境調查與效益評估,” 國立高雄海事專科學校漁業科.
[7] 黃材成, ”人工魚礁校慶展示實況—海洋培育復育之新利器,” 國立中山大學
海洋環境及工程學系全球資訊網首頁, 1999.
[8] 劉春成, ”臺灣沿岸漁業及漁場環境資料庫,” 86科技-1.12-漁-04(10), 行政院
農業委員會, 1997.
[9] 羅聖宗與陳民本, “海床沈積物的採樣,” 科儀新知, 第十二卷第六期, 1991.
[10] 劉金源,水中聲學—原理與應用, 國立中山大學出版社, 1999.
[11] 陳屏先, 海洋聲學傳播模組分析, 國立臺灣大學造船及海洋工程學研究所
碩士論文, 1998.
[12] 陳民本, 宋國士, 徐春田與吳軍梁, 台灣四周海床底質資料庫, 行政院國
家科學委員會國防科技研究報告, 1992.
[13] Shumway, 1960; Hamilton, 1965; Buchan et al., 1967; Hamilton, 1979; Mayer,
1979; Tucholke, 1980; Hamilton et al., 1982; Bachman, 1985; Badiey et al.,
1988; Brunson and Johnson, 1980; Davis et al., 1989; Van Weering et al., 1989.
[14] 唐存勇, 黃哲崇, 林斐然與劉康克, “國科會KEEP(Kuroshio-Edge Exchange
Processes)計劃黑潮陸棚邊緣交換過程研究,” 國科會KEEP計畫資料庫系統,
1996.
[15] J. C. Hassab, Underwater Signal and Data Processing, CRC Press, 1989.
[16] R. J. Urick, Principles of Underwater Sound, 3rd ed., McGraw-Hill, New York,
1983.
[17] A. V. Oppenheim, Application of Digital Signal Processing, Prentice Hall, l977.


[18] A. V. Oppenheim and R. W. Schafer, Discrete-Time Signal Processing, Prentice
Hall, l989.
[19] W. S. Burdic, Underwater Acoustic System Analysis, N. J. Prentice Hall, 1984.
[20] W. C. Knight, “Digital Signal Prcessing for Sonar,” Proc. of IEEE(The Institute
of Electrical and Electronics Engineers, Inc.), pp. 1451-1506, Nov. 1981.
[21] F. L. Casselman, D. F. Freeman, D. A. Kerrigan, S. E. Lane, N. H. Millstrom,
and W. G. Nichols Jr., “A Neural Network-Based Passive Sonar Detection and
Classification Design with a Low False Alarm Rate,” GTE Government Systems
Command, Control and Communications Systems, 77 “A” Street Needham, MA
02194.
[22] R. P. Gorman and T. J. Sejnowski, “Learned Classification of Sonar Targets
Using a Massively Parallel Network,” IEEE Transactions on Acoustics, Speech,
and Signal Processing, Vol. 36, No. 7, July 1988.
[23] D. W. Cottle and D. J. Hamilton, “All Neural Network Sonar Discrimination
System,” Raytheon Submarine Signal Division Portsmouth, RI(Rhode Island).
[24] D. E. Rumelhart and T. Sejnowski, “Neural Network Architectures for AI,”
American Association for Artificial Intelligence, 11th International Conference on
Artificial Intelligence held at Detroit, Michigan, August 1989.
[25] T. B. Haley, “Applying Neural Networks to Automatic Active Sonar
Classification,” IEEE l0th International Conference on Pattern Recognition, Vol.
2, pp. 41-44, 1990.
[26] 黃益良, 類神經網路應用於聲源訊號之分類研究, 私立中原大學資訊工程
研究所碩士論文, 1996.
[27] 沈清與湯霖, 模式識別導論, 中華人民共和國國防科技大學出版社, 1993.
[28] 黃皇嘉, 類神經網路應用於即時聲源訊號辨識系統, 私立中原大學資訊工
程研究所碩士論文, 1997.
[29] 蔡瑞煌, 類神經網路概論, 三民書局, 1995.
[30] 海研三號研究船目錄, 國立中山大學海洋科學院海研三號船務室, 1988.
[31] 海洋研究船貴重儀器使用中心網頁, 國家海洋科學研究中心, 1999.
[32] 童乃儉, “簡介多波束測深儀摘要,” 海下技術季刊八十五年度第六卷第一期,
中華民國海下技術協會, 1996.
[33] 擎天信使聲集團GTXS網頁, 擎天信使音樂製作有限公司, 1999.
[34] Portable DAT Instrumentation Recorder, “PC200Ax Series,” Sony Magnescale
Inc., May 1995.
[35] 徐學群,孫郁興與蘇順吉, “建構於個人電腦平台上之水下聲紋辨識系統,”
國際海洋年「海洋，海軍，科技」研討會論文集, 1998.


[36] W. A. Gardner, Statistical Spectral Analysis, Prentice Hall, pp.3-27, 34-104, 179-
240, 1988.
[37] Papoulis, Signal Analysis, McGraw-Hill, New York, 1977.
[38] J. Wolcin, “On the Statistical Properties of Noise Background Equalization
Schemes,” NCSC Tech. Memo. No. 781159, Naval Underwater System Center,
New London, Ct-31 July 1978.
[39] W. Struzinski, “A New Normalization Algorithm for Detection Systems,” J.
Acoust. Soc. Am. Suppl. 175, S43, 1984.
[40] William A. Struzinski and Edward D. Lowe, “A Performance Comparison of
Four Noise Background Normalization Schemes Proposed for Signal Detection
Systems,” J. Acoust. Soc. Am. Vol. 76, No. 6, Dec. 1984.
[41] Merrill I. Skolnik, Radar Applications, McGraw-Hill, New York, 1980.
[42] 焦李成, 神經網路系統理論, 儒林圖書有限公司, 1991.
[43] Michael Chester, Neutral Networks: a tutorial, PTR Prentice Hall, 1993.
[44] Phil Mars, J. R. Chen, and Raghu Nambiar, Learning Algorithms: Theory and
Applications in Signal Processing, Control and Communications, CRC Press,
1996.
[45] 詹文欽,類神經網路容錯能力及快速學習之研究, 國立中山大學電機工程研
究所碩士論文, 1993.
[46] 曹大鵬, “類神經網路在馬達聲訊辨識上之應用,” 行政院國家科學委員會
專題研究計畫成果報告, 1997.
[47] W. S. McCulloch, and W. Pitts, “A Logical Calculus of the Ideas Immanent in
Nervous Activity,” Bulletin of Math. Bio. , Vol. 5, pp. 115-133, 1943.
[48] Martin T. Hagan, Howard B. Demuth, and Mark Beale, Neutral Network Design,
PWS Publishing Company, 1996.
[49] 劉省宏, 類神經網路應用在上肢動作描繪和力矩值估計, 國立成功大學醫
學工程研究所碩士論文, 1992.
[50] 章定遠, 貝氏類神經網路及HIDDEN MARKOV MODEL應用於上肢肌電信
號之辨認, 國立成功大學醫學工程研究所碩士論文, 1992.
[51] R. Hecht-Nielsen, Neurocomputing, Wesley Publishing Company, New York,
1996.
[52] 葉怡成, 類神經網路模式應用與實作, 儒林圖書有限公司, 1999.
[53] Julius S. Bendat, and Allan G. Piersol, Random Data: Analysis and Measurement
Procedures, John Wiley & Sons, 1986.
[54] James Meister, “A Neural Network Harmonic Family Classifier,” J. Acoust. Soc.
Am. 93(5), pp. 1488-1495, 1993.


[55] W. Chang, B. Bosworth, and G. Clifford Carter, “Results of Using an Artificial
Neural Network to Distinguish Single Echoes from Multiple Sonar Echoes,” J.
Acoust. Soc. Am. 94(3), pp. 1404-1408, 1993.
[56] Ronald J. Gibbs, Suspended Solids in Water, Plenum Press, New York, 1974.
[57] W. Miller, T. McKenna, and C. Lau, “Office of Naval Research Contributions to
Neutral Networks and Signal Processing in Oceanic Engineering,” IEEE J.
Ocean Eng. 17(4), pp. 299-307, 1992.
[58] W. Chang, B. Bosworth, and G. Clifford Carter, “Letters: Empirical Results of
Using Back-propagation Neural Networks to Separate Single Echoes from
Multiple Echoes,” IEEE Trans. Neural Networks, vol. 4, no. 6, pp. 993-995,
1993.
[59] B. W. McCleave, J. K. Owens, and Frank M. Ingels, “Analyzing Depth Sounder
Signals with Artificial Neural Networks,” Sea technology 92(march), pp. 39-42,
1992.
[60] Michael Wroth, “Survey of Neural Network Technology for Automatic Target
Recognition,” IEEE Trans Neural Networks, vol. 1, no. 1, pp. 28-43, 1990.
[61] Richard P. Lippmann, “Pattern Classification Using Neural Networks,” IEEE
Commun. Mag., pp. 47-64, Nov. 1989.
[62] Anders Svardstrom, “Neural Network Feature Vectors for Sonar Targets
Classifications,” J. Acoust. Soc. Am. 93(5), pp. 2656-2665, 1993.
[63] W. C. Krumbein, and L. L. Sloss, “Stratigraphy and Sedimentation,” W. H.
Freeman and Company, San Francisco, 1963.
[64] William T. Collins, and P. Lacroix, “Operational Philosophy of Acoustic
Waveform Data Processing for Seabed Classification,” Oceanology International
97 Pacific Rim, Singapore, May 1998.
[65] 林銘祺,類神經網路於水中聲響訊號分類的應用, 國立成功大學電機工程研
究所碩士論文, 1992.
[66] William T. Collins, and J. L. Galloway, “Seabed Classification with Multibeam
Bathymetry,” Sea technology 98(September), pp. 45-49, 1998.
[67] Biosonics, “Bottom Classification Detailed,” Biosonics全球資訊網首頁, 1999.
[68] Lester R. LeBlanc, Larry Mayer, Manuel Rufino, Steven G. Schock, and John
King, "Marine Sediment Classification Using the Chirp Sonar," J. Acoust. Soc.
Am. 91(1), pp. 107-115, January 1992.