臺灣博碩士論文加值系統

自從類神經網路被發明以來，其中的多層感知器被應用的範圍是最廣亦最為熱門。但多層感知器(Multi-Layer Perceptron，MLP)本身隱藏層類神經元的架構過於簡單，而有收斂速度過慢等問題。本篇論文將提出以分流式抑制類神經網路(Shunting Inhibitory Neural Networks，SIANN)及一般化前饋式類神經網路(Generalized Feedforward Neural Networks，GFNN)配合傳統的倒傳遞演算法(Back-Propagation Algorithm)來進行手寫數字辨識的研究，並與傳統的多層感知器進行相互間的比較。我們採用了MNIST數字資料庫當成辨識與測試的輸入資料來源，本文尚有文字切割、正規化等介紹，並以方向性距離分佈(Directional Distance Distribution，DDD)作為特徵抽取的方式，可以表現字型的結構性及整體性，達到高準確的辨識率。經實驗證明，SIANN雖然辨識率比MLP要差，但其收斂速度優於MLP；而GFNN的辨識率及收斂速度都優於MLP及SIANN，可達到98.4%的辨識率。

目錄……………………………………………………………………I
圖目錄、表目錄 …………………………………………………… III
第一章 緒論
1.1 研究動機………………………………………………… 1
1.2 離線手寫辨識簡介..…………………………………… 3
1.3 文獻探討………………………………………………… 8
1.4 論文架構………………………………………………… 13
1.5 辨識系統流程….……………………………………… 14
第二章 前處理及特徵抽取法則
2.1 MNIST手寫數字資料庫.………………………………… 16
2.2 影像辨識前置處理……………………………………… 17
2.2.1 空白切割………………………………………..... 17
2.2.2 字元正規化……………………………………..... 19
2.3 特徵抽取法則…….……………………………………….. 21
2.3.1 方向性距離分佈特徵介紹………………………. 22
2.3.2 貼圖………………………………………………. 24
2.3.3 方向性距離分佈特徵抽取法則…………………. 25
第三章 MLP、SIANN與GFNN類神經網路
3.1 類神經網路簡介………………………………………….. 30
3.1.1 類神經網路分類………………………………… 35
3.2 多層感知器……....……………………………………….. 37
3.2.1 MLP配合倒傳遞演算法………………………… 41
3.2.2 倒傳遞演算法參數設定…………………………. 47
3.3 分流式抑制類神經網路………………………………….. 50
3.3.1 SIANN配合倒傳遞演算法……………………… 53
3.3.2 SIANN架構探討………………………………… 57
3.4 一般化前饋式類神經網路….…………………………….. 58
3.4.1 GFNN配合倒傳遞演算法………………………. 60
3.4.2 GFNN探討………………………………………. 65
3.5 XOR問題………………...……………………………….. 65
第四章 模擬結果與討論
4.1 實驗方法………………………………………………….. 68
4.2 分析與討論……………………………………………….. 70
4.3 實驗結果與比較………………………………………….. 96
4.4 辨識結果比較…………………………………………….. 97
第五章 結論與未來研究方向
5.1 結論………………….……………………………….…… 98
5.2 未來研究方向…………………………………………….. 98
參考文獻…………………………………………………………… 100

[1]Ling Gang, Verma B. and Kulkami S., ” Experimental analysis of neural network based feature extractors for cursive handwriting recognition”, Neural Networks, 2002. IJCNN '02. Proceedings of the 2002 International Joint Conference on , Volume: 3 , 12-17 May 2002
Pages:2837 – 2842
[2]Arulampalam, G. and Bouzerdoum, A., ” Application of shunting inhibitory artificial neural networks to medical diagnosis”, Intelligent Information Systems Conference, The Seventh Australian and New Zealand 2001 , 18-21 Nov. 2001 Pages:89 - 94
[3]Bouzerdoum, A., ” A new class of high-order neural networks with nonlinear decision boundaries”, Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on , Volume: 3 , 16-20 Nov. 1999 Pages:1004 - 1009 vol.3 Pages:2837 – 2842
[4]Tivive, F.H.C. and Bouzerdoum, A., ” A new class of convolutional neural networks (SICoNNets) and their application of face detection”, Neural Networks, 2003. Proceedings of the International Joint Conference on , Volume: 3 , 20-24 July 2003 Pages:2157 - 2162 vol.3
[5]Hammadou, T. and Bouzerdoum, A., ” Novel image enhancement technique using shunting inhibitory cellular neural networks”, Consumer Electronics, 2001. ICCE. International Conference on , 19-21 June 2001 Pages:284 - 285
[6]Arulampalam, G. and Bouzerdoum, A., ” A generalized feedforward neural network classifier”, Neural Networks, 2003. Proceedings of the International Joint Conference on , Volume: 2 , 20-24 July 2003
Pages:1429 - 1434 vol.2
[7]Il.-Seok. Oh and Ching Y. Suen., ” Distance features for neural network-based recognition of handwritten characters”, IJDAR, 1(2):73--88, 1998.
[8]Bin Zhang and Srihari, S.N., ” Fast k-nearest neighbor classification using cluster-based trees”, Pattern Analysis and Machine Intelligence, IEEE Transactions on , Volume: 26 , Issue: 4 , April 2004 Pages:525 - 528
[9]Qing Wang, Rongchun Zhao, Zheru Chi and Feng, D.D., ” HMMRF: a stochastic model for offline handwritten Chinese character recognition”, Signal Processing Proceedings, 2000. WCCC-ICSP 2000. 5th International Conference on , Volume: 3 , 21-25 Aug. 2000
Pages:1475 - 1478 vol.3
[10]Wen Jin, Zhao Jia Li, Luo Si Wei and Han Zhen, ” The improvements of BP neural network learning algorithm”, Signal Processing Proceedings, 2000. WCCC-ICSP 2000. 5th International Conference on , Volume: 3 , 21-25 Aug. 2000 Pages:1647 - 1649 vol.3
[11]Correia, S.E.N., de Carvalho, J.M. and Sabourin, R., ” On the performance of wavelets for handwritten numerals recognition”, Pattern Recognition, 2002. Proceedings. 16th International Conference on , Volume: 3 , 11-15 Aug. 2002 Pages:127 - 130 vol.3
[12]Verma, B., ”A feature extraction technique in conjunction with neural network to classify cursive segmented handwritten characters”, Neural Networks Proceedings, 1998. IEEE World Congress on Computational Intelligence. The 1998 IEEE International Joint Conference on , Volume: 1 , 4-9 May 1998
Pages:332 - 336 vol.1
[13]Ohkura, M., Shimada, Y., Shiono, M. and Hashimoto, R., ” On discrimination of handwritten similar KANJI characters by subspace method using several features”, Document Analysis and Recognition, 1993., Proceedings of the Second International Conference on , 20-22 Oct. 1993 Pages:589 - 592
[14]Yamada, H., Yamamoto, K. and Saito, T., ” A nonlinear normalization method for handprinted kanji character recognition -line density equalization”, Pattern Recognition, 1988., 9th International Conference on , 14-17 Nov. 1988 Pages:172 - 175 vol.1
[15]Qing Wang, Zheru Chi, Feng, D.D. and Rongchun Zhao, ” Match between normalization schemes and feature sets for handwritten Chinese character recognition”, Document Analysis and Recognition, 2001. Proceedings. Sixth International Conference on , 10-13 Sept. 2001 Pages:551 – 555
[16]Cheung, H.N., Bouzerdoum, A. and Newland, W., ” Properties of shunting inhibitory cellular neural networks for colour image enhancement”, Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on , Volume: 3 , 16-20 Nov. 1999 Pages:1219 - 1223 vol.3
[17]Bouzerdoum, A. and Pinter, R.B., ” A shunting inhibitory motion detector that can account for the functional characteristics of fly motion-sensitive interneurons”, Neural Networks, 1990., 1990 IJCNN International Joint Conference on , 17-21 June 1990 Pages:149 - 153 vol.1
[18]Arulampalam, G. and Bouzerdoum, A., ” Expanding the structure of shunting inhibitory artificial neural network classifiers”, Neural Networks, 2002. IJCNN '02. Proceedings of the 2002 International Joint Conference on , Volume: 3 , 12-17 May 2002 Pages:2855 – 2860
[19]Bhattacharya, U. and Chaudhuri, B.B., ” A majority voting scheme for multiresolution recognition of handprinted numerals”, Document Analysis and Recognition, 2003. Proceedings. Seventh International Conference on , 3-6 Aug. 2003 Pages:16 - 20 vol.1
[20]魏銪志, 動態多標的車牌辨識系統之研究, 元智大學資訊管理研究所碩士論文, 2000
[21]陳嘉雄, 多角度人臉之二階段辨識, 國立東華大學資訊工程
學系碩士論文, 2002
[22]林穎駿, 應用小波轉換與類神經網路於電力品質事故之監測與辨識, 中原大學電機工程研究所碩士論文, 2003
[23]李健宏, 植基於WGLVQ離線式手寫數字辨識, 台灣師範大學工業教育學系在職進修碩士論文, 2003
[24]陳毓山, 螞蟻演算法最佳化倒傳遞類神經網路於土層剪力波速評估之研究, 國立台灣大學土木工程學研究所碩士論文, 2003
[25]林瑞明, 手寫紙本是非及選擇題答案辨識系統之實作, 台中師範學院教育測驗統計研究所碩士論文, 2002
[26]郭癸蘭, 手寫身份證字號辨識系統, 高雄第一科技大學電腦與通訊工程系碩士論文, 2002
[27]呂信德, 利用多重專家之車牌辨識系統, 國立中央大學資訊工程研究所碩士論文, 2002
[28]蔡明志, 神經網路應用於字元的不變性辨識, 元智大學工業工程研究所碩士論文, 2000
[29]溫福助, 類神經網路樣板比對法於車牌字元辨識之研究, 國立台灣大學電機工程學研究所碩士論文, 2000
[30]繆紹綱, 數位影像處理, 第十章, 全華科技圖書股份有限公司, 1999
[31]陳德, 模糊類神經網路結合進化演算法運用在基頻通道等化器上, 國立中央大學電機工程研究所碩士論文, 2001
[32]葉怡成, 類神經網路模式應用與實作, 儒林圖書公司, 2003