臺灣博碩士論文加值系統

本論文提出了一種新穎的嘴唇辨識方法利用活動基模形(Active Basis Model, ABM)。此方法的流程架構主要包含四個階段。在第一階段，取得嘴唇圖形的變形樣板。這些製作變形樣板的嘴唇圖形為明顯開嘴或明顯閉嘴。第二階段，取得測試圖形的基礎樣板。第三階段，比較訓練出的變形樣板與每張圖片的基礎樣間的概似性，並作為特徵向量。最後，利用支撐向量機（Support Vector Machine，SVM）分類第三階段的向量特徵。在本文中使用支撐向量機分類嘴唇是一種監督式的學習。本文從BioID與PCIS人臉資料庫中擷取出嘴唇圖片作為實驗資料庫，準確率最高可達96.375％。實驗結果證明本文提出的方法對於嘴唇影像識別擁有極佳的辨識結果。嘴唇的辨識结果可用於自動語音識別與臉部辨識系統。

The paper proposed a novel method for lip recognition based on Active Basis Model (ABM). There are four stages in a flowchart of this novel method. At the first stage, the deformable templates of lip images is obtained. The lip images of deformable templates are obvious open or closed. The second stage is to obtain the deformed template of each testing images. The third stage, the difference between the deformable template and deformed template is calculated and used as a feature vector. Finally, the support vector machine (SVM) is use to classify the feature vector. SVM is a supervised learning to be a classifier for lip recognition. Experimental results are used to verify the proposed method has good performance. There are 1000 face images in BioID and PCIS face database used for the experiment. The most accuracy of results is 96.375%. Lips classification results can be used in automatic speech recognition and face recognition system.

目錄
摘要........................................................................................................................I
ABSTRACT ........................................................................................................ II
圖目錄.................................................................................................................VI
表目錄.............................................................................................................. VIII
第1 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................................................................................................. 3
第2 章 相關研究................................................................................................. 5
第3 章 特徵擷取................................................................................................. 8
3.1 賈巴濾波.................................................................................................. 8
3.2 活動基模型............................................................................................ 10
3.3 共同勾畫演算法(shared sketch algorithm)........................................ 12
3.4 向量特徵............................................................................................... 14
第4 章 支撐向量機........................................................................................... 15
4.1 線性不可容錯........................................................................................ 15
4.2 線性可容錯............................................................................................ 17
4.3 核函數.................................................................................................... 19
4.4 SVM 開發工具...................................................................................... 20
第5 章 實驗結果與討論................................................................................... 27
第6 章 結論與未來工作................................................................................... 37
參考文獻............................................................................................................. 38
圖目錄
圖3-1 8 個方向、5 種縮放大小，40 組賈巴濾波組........................................ 9
圖3-2 影像旋積後結果..................................................................................... 10
圖3-3 活動基模型............................................................................................. 11
圖3-4 變型樣板與基礎樣板............................................................................. 12
圖3-5 元素選擇................................................................................................. 13
圖4-1 二維平面線性可分SVM 分類示意圖.................................................. 17
圖4-2 二維平面線性不可分SVM 分類示意圖.............................................. 19
圖4-3 映射函數示意圖...................................................................................... 20
圖4-4 LIBSVM 在Window 平台執行結果...................................................... 21
圖4-5 svmtoy 執行畫面..................................................................................... 24
圖4-6 svmtoy 佈點畫面..................................................................................... 24
圖4-7 svmtoy 分類畫面..................................................................................... 24
圖4-8 svmtoy 修改參數結果............................................................................. 25
圖4-9 LIBSVM 暴力尋找參數......................................................................... 25
圖5-1 BioID 人臉資料庫示意圖...................................................................... 28
圖5-2 PCIS 人臉資料庫示意圖........................................................................ 29
圖5-3 容易導致誤判圖片................................................................................. 30
圖5-4 縮放係數s=1.0，各樣本準確率........................................................... 33
圖5-5 縮放係數s=1.0，基元數量修正後，各樣本準確率.......................... 34
圖5-6 基元數量40，各樣本準確率............................................................... 35
圖5-7 基元數量40，縮放系數修正後，各樣本準確率............................... 36
表目錄
表5-1 不同基元數量參照不同縮放係數樣板................................................ 32
表5-2 數據表對於圖5-4 ................................................................................... 33
表5-3 數據表對於圖5-5 ................................................................................... 34
表5-4 數據表對於圖5-6 ................................................................................... 35
表5-5 數據表對於圖5-7 ................................................................................... 36

[1]陳建霖，以Java為平台建構布袋戲網路戲院，碩士論文，朝陽科技大學資訊與通訊系，台中，2009。
[2]Liou, C.Y ., (1996), “Active mesh for lip-reading,” http://ntur.lib.ntu.edu.tw/bitstream/246246/8434/1/882213E002036.pdf
[3]Hao, T., F. Yun, J. Tu, S. Thomas .S. Huang, and Mark Hasegawa-Johnson, (2008), “EAVA: A 3D Emotive Audio-Visual Avatar,” 2008 IEEE Workshop on Applications of Computer Vision (WACV''08), pp.1-6.
[4]Hao, T, F. Yun, J. Tu, Mark Hasegawa-Johnson, and Thomas S. Huang, (2008), “Humanoid Audio-Visual Avatar with Emotive Text-To-Speech Synthesis,” IEEE Transactions on Multimedia (T-MM), Vol. 10, No. 6, pp. 969-981.
[5]Ju, J.S., Shin, Y.H., Kim, E.Y., (2009),” Vision based interface system for hands free control of an intelligent Wheelchair,” Journal of Neuro Engineering and Rehabilitation.
[6]W. S. Yanbor, B. A. Draper, J. R. Beveridge, (2002), “Analyzing PCA-based Face Recognition Algorithms: Eigenvector Selection and Distance Measures,” In Empirical Evaluation Methods in Computer Vision. World Scientific Press.
[7]Z. Jin, Z. Lou, J. Yang, and Q. Sun, (2007), “Face detection using template matching and skin-color information,” Neurocomputing, vol. 70, no. 4-6, pp. 794-800.
[8]Nikolaidis, A. and I. Pitas, (2000), “Facial feature extraction and pose determination,” Pattern Recognition, vol. 33, pp. 1783-1791.
[9]I. Kotsia and I. Pitas, (2007), “Facial Expression Recognition in Image Sequences Using Geometric Deformation Features and Support Vector Machines,” IEEE Transactions on Image Processing, Vol. 16, pp. 127-187, January 2007.
[10]D. H. Kim, S. U. Jung and M. J. Chung, (2008), “Extension of cascaded simple feature based face detection to facial expression recognition,” Pattern Recognition Letters, vol.29, pp. 1621-1631, August 2008.
[11]Peng Yanga, Qingshan Liua and Dimitris N. Metaxas, (2009), “Boosting encoded dynamic features for facial expression recognition,” Pattern Recognition Letters, vol.30, pp.132-139, January 2009.
[12]P. Viola and M. Jones, (2002), “Robust real-time object detection,” International Journal of Computer Vision.
[13]Wu, Y.N., Si, Z.Z., Fleming, C., Zhu, S.C., (2007), “Deformable Template As Active Basis,” Proceedings of International Conference on Computer Vision.
[14]Wu, Y. N., Si, Z.Z., Gong. H., Zhu, S.C., (2010), “Learning active basis model for object detectionand recognition,” International Journal of Computer Vision, in press.
[15]Si, Z. Z., Wu, Y. N., 2010, “Wavelet, active basis, and shape script,” http://www.stat.ucla.edu/~ywu/AB/mir311sFeb2010.pdf.
[16]Si, Z. Z., Gong, H., Zhu , S. C., Wu, Y. N., (2010), “Learning active basis models by EM-type algorithms,” Statistical Science, in press.
[17]Hui Tang,(2009), “Comparison of an active basis model and an adaboost model in image representation,” http://theses.stat.ucla.edu/105/Hui%20thesis.pdf
[18]D.Gabor, (1946), “Theory of Communication”, J.Inst. Elec. Eng., vol. 93, 429-459.
[19]J. G. Daugman, (1985), “Uncertainty relation for resolution in space, spatial frequency, and orientation optimized by two-dimensional visual cortical filters,” Journal of the Optical Society of America, vol.2, no. 7, pp. 1160-1169.
[20]Sheng-Wen Wang, “Content_Based Image Retrievals Using Various Texture Features,” Master Thesis, Department of Computer Science, University of National Tsing Hua, Hsinchu, 2001.
[21]Kedar A. Patwardhan, Guillermo Sapiro, Vassilios Morellas (2008). “Robust Foreground Detection in Video Using Pixel Layers,” IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, vol. 30, no. 4, 746-751
[22]F. Qi, C. Bao, and Y. Liu (2004). “A novel two-step SVM classifier for voiced/ unvoiced/ silence classification of speech,” Proc. Int. Symp. Chinese Spoken Language, 77-80.
[23]J. Y. Lai, A. Sowmya, and J. Trinder (2005). “Support vector machine experiments for road recognition in high resolution images,” Lecture Notes in Artificial Intelligence, 426-435.
[24]Cortes and V. Vapnik, “Support-vector networks,” Machine Learning, vol. 20, pp. 273-297, 1995.
[25]B. E., Boser, I. M. Guyon, and V. N. Vapnik, (1992), “A training algorithm for optimal margin classifiers,” Proceedings of the Fifth Annual Workshop on Computational Learning Theory, Pittsburgh, Pennsylvania, United States, pp.144-152.
[26]C. C. Chang and C. J. Lin, (2001),“LIBSVM:a library for support vector machines,”, Software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm