請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/27940
完整後設資料紀錄
DC 欄位 | 值 | 語言 |
---|---|---|
dc.contributor.advisor | 丁建均(Jian-Jiun Ding) | |
dc.contributor.author | Wei-Lun Chao | en |
dc.contributor.author | 趙偉崙 | zh_TW |
dc.date.accessioned | 2021-06-12T18:29:08Z | - |
dc.date.available | 2013-09-08 | |
dc.date.copyright | 2011-09-08 | |
dc.date.issued | 2011 | |
dc.date.submitted | 2011-08-08 | |
dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/27940 | - |
dc.description.abstract | 基於可接觸資料的增加與計算技術的快速發展,過去十年,機器學習已因為人類生活中大量的自動化需求而吸引了許多注意力。現今在物件識別、機器人學、人工智慧、電腦視覺、甚至是經濟學等科學領域當中,機器學習已成為從資料當中抽取與探索重要資訊不可或缺的角色。
另一方面,在過去幾十年,人臉偵測與辨識等人臉相關的主題已漸漸成為物件識別與電腦視覺中的重要研究領域。其原因來自於自動化識別與監視系統的需求、對於人類視覺系統在人臉感知上的興趣、與人機互動介面的設計開發等。 在本篇論文當中,我們專注在使用機器學習技術來估測人類的年齡。我們提出了一個人類年齡估測的架構,其包含了特徵抽取、距離度量調整、降低維度、與年齡測定等四個步驟。我們使用了相當受歡迎的主動外觀模型(active appearance model, AAM)來抽取特徵。此特徵可以共同地表述人臉形狀與質地的變化。為了增強監督式降維演算法的效果,我們使用相關組件分析(relevant component analysis, RCA)技術來達成距離度量的調整。因為年齡標籤本身具有順序性的關係,我們提出了一個標籤敏感(label-sensitive)的概念,以更有效的在距離度量與降維學習的過程當中運用標籤的資訊。此外,我們還提出了一些修改來減輕資料庫當中可能存在的不平衡問題。 根據降維過程中保存局部或鄰近資訊的特性,我們使用局部性回歸而非總體性回歸來測定年齡。由實驗結果中與其它方法的比較,我們所提出的架構與修改在最普遍使用的FG-Net資料庫中可以達到最低的平均絕對誤差。 | zh_TW |
dc.description.abstract | Based on the increasing of accessible data and the fast development of the computational technology, machine learning attracted lots of attention in the last ten years because of the great demand of automation in human life. Now in the disciplines of pattern recognition, robotics, artificial intelligences, computer vision, and even economics, machine learning has been an indispensible part to extract and discover the valuable information from data.
On the other hand, human face related topics such as face detection and recognition became important research fields in pattern recognition and computer vision during the last few decades. This is due to the needs of automatic recognition and the surveillance system, the interest in the human visual system on human face perception, and the design of human-computer interface, etc. In this thesis, we focus on using machine learning techniques for human age estimation. A human age estimation framework is proposed, which includes four steps: feature extraction, distance metric adjustment, dimensionality reduction, and age determination. The popular active appearance model (AAM) is exploited in our work for feature extraction, which can jointly represent the shape and texture variations of human faces. To enhance the performance of supervised dimensionality reduction, the relevant component analysis (RCA) is used for distance metric adjustment. Because human ages are with ordinal relationship, we proposed the “label-sensitive” concept to better exploit aging labels during distance metric learning and dimensionality reduction learning. In addition, several modifications are proposed to alleviate the possible unbalance problem in the database. According to the neighbor / locality preserving characteristic during dimensionality reduction, the proposed framework utilizes local regression instead of global regression for age determination. From the experimental results, the proposed framework and its modification versions achieve the lowest mean absolute error (MAE) over other existing techniques in the most widely-used FG-Net database, both in the LOPO and cross-validation settings. | en |
dc.description.provenance | Made available in DSpace on 2021-06-12T18:29:08Z (GMT). No. of bitstreams: 1 ntu-100-R98942073-1.pdf: 6633536 bytes, checksum: 017e3b4e679c8028d9823b86748a3596 (MD5) Previous issue date: 2011 | en |
dc.description.tableofcontents | 口試委員會審定書 #
誌謝 i 中文摘要 iii ABSTRACT v CONTENTS vii LIST OF FIGURES xv LIST OF TABLES xxvii Chapter 1 Introduction 1 1.1 Motivation 1 1.2 Main Contribution 1 1.3 Organization 2 1.4 Notation 3 Chapter 2 Machine Learning Techniques Overview 4 2.1 What is Machine Learning? 5 2.1.1 Notation of Dataset 5 2.1.2 Training Set and Test Set 6 2.1.3 No Free Lunch Rules 7 2.1.4 Relationships with Other Disciplines 10 2.2 Basic Concepts and Ideals of Machine Learning 10 2.2.1 Designing versus Learning 11 2.2.2 The Categorization of Machine Learning 11 2.2.3 The Structure of Learning 14 2.2.4 What are We Seeking? 16 2.2.5 The Optimization Criterion for Supervised Learning 19 2.2.6 The Strategies of Supervised Learning 25 2.3 Principles and Effects of Machine Learning 28 2.3.1 The VC Bound and Generalization Error 29 2.3.2 Three Learning Effects 31 2.3.3 Feature Transform 38 2.3.4 Model Selection 40 2.3.5 Three Learning Principles 44 2.3.6 Practical Usage: A First Glance 45 2.4 Techniques of Supervised Learning 47 2.4.1 Supervised Learning Overview 47 2.4.2 Linear Model (Numerical Functions) 50 2.4.3 Conclusion and Summary 64 2.5 Techniques of Unsupervised Learning 64 2.6 Practical Usage: Pattern Recognition 65 2.7 Conclusion 67 Chapter 3 Dimensionality Reduction 69 3.1 Why do We Need Dimensionality Reduction? 70 3.1.1 Practical Reasons 71 3.1.2 Theoretical Reasons 73 3.2 Dimensionality Reduction Overview 75 3.2.1 Strategies of Dimensionality Reduction 76 3.2.2 Topology and Embedding 77 3.2.3 Categorizations of Dimensionality Reduction 78 3.3 Principal Component Analysis (PCA) 81 3.4 Linear Discriminant Analysis (LDA) 86 3.5 Multidimensional Scaling (MDS) 89 3.6 Isomap and Its Extensions 94 3.6.1 Isomap: Isometric Feature Mapping 95 3.6.2 Extensions of Isomap 99 3.7 Locally Linear Embedding and Its Extension 101 3.7.1 Locally Linear Embedding (LLE) 101 3.7.2 Extensions of LLE 108 3.8 Laplacian Eigenmap and Its Extensions 110 3.8.1 Laplacian Eigenmap 110 3.8.2 Extensions of Laplacian Eigenmap 113 3.9 Graph Embedding and Its Extensions 117 3.9.1 The Graph Embedding Framework 117 3.9.2 Marginal Fisher Analysis 121 3.9.3 Extensions and Related Work of Graph Embedding 123 3.10 Other Nonlinear Dimensionality Reduction Techniques 126 3.10.1 Manifold Alignment 126 3.10.2 Maximum Variance Unfolding 128 3.10.3 Kernel PCA and Kernel LDA 129 3.10.4 Independent Component Analysis 129 3.10.5 Other Methods 130 3.11 Feature Selection 130 3.12 Conclusion 131 Chapter 4 Face Detection and Recognition 133 4.1 Introduction to Face Detection and Face Recognition 134 4.1.1 Face Detection 134 4.1.2 Feature Extraction 134 4.1.3 Face Recognition 135 4.2 Issues and Factor of Human Face 136 4.2.1 Domain Knowledge of Human Face and Human Visual Perception 136 4.2.2 Factors of Human Appearance Variations 138 4.2.3 Design Issues 141 4.3 Face Detection 142 4.3.1 Knowledge-Based Face Detection 142 4.3.2 Feature Invariant Approaches 144 4.3.3 Template Matching Methods: Active Appearance Model 149 4.3.4 Appearance-Based Methods 154 4.3.5 Part-Based Methods 158 4.4 Feature Extraction and Face Recognition 163 4.4.1 Holistic-based methods 164 4.4.2 Feature-based Methods 174 4.4.3 Template-based Methods 182 4.4.4 Part-based Methods 183 4.5 Conclusion 188 Chapter 5 Facial Age Estimation 189 5.1 Challenges of Facial Age Estimation 190 5.1.1 Intrinsic Factors 190 5.1.2 Technical Factors 191 5.2 The General Framework 192 5.2.1 The Algorithm Structure of Facial Age Estimation 193 5.2.2 The Performance Evaluation Criterion 196 5.2.3 The Face Age Databases 196 5.3 Facial Feature Extraction 197 5.3.1 Anthropometric Model 197 5.3.2 Active Appearance Model (AAM) 198 5.3.3 Aging Pattern Subspace (AGES) 199 5.3.4 Bio-Inspired Features (BIF) 202 5.3.5 Aging Manifold 204 5.3.6 Spatially Flexible Patches (SFP) 205 5.3.7 Gabor Features 208 5.3.8 Local Binary Patterns 208 5.3.9 Other Kinds of Feature Extraction Methods 209 5.4 Age determination 209 5.4.1 Multi-class Classification for Age Determination 210 5.4.2 Regressor for Age Determination 212 5.4.3 Hybrid Combination 215 5.4.4 Ordinal Regression 221 5.4.5 Cost-Sensitive Classification 224 5.4.6 Other Methods 226 5.5 Experimental Settings 226 5.6 Conclusion 227 Chapter 6 Proposed Methods and Modifications 229 6.1 The Proposed Framework and Important Factors 229 6.1.1 The Proposed Framework 230 6.1.2 Important Factors 231 6.1.3 The Adopted Database 232 6.1.4 The Adopted Feature Extraction Method 233 6.2 The modified Distance Metric Learning Algorithm 235 6.2.1 A Brief Overview of Distance Metric Learning 236 6.2.2 The Proposed Usage of Distance Metric Learning Algorithm 237 6.2.3 The Proposed Modification on RCA 239 6.2.4 Simulation Results 241 6.3 The Proposed Dimensionality Reduction Algorithm 246 6.3.1 Extending Dimensionality Reduction for Ordering Labels 246 6.3.2 Label-Sensitive Graph Embedding (lsGE) 247 6.3.3 Label-Sensitive Locality Preserving Projection (lsLPP) 251 6.3.4 The Modifications for Unbalanced Data 252 6.3.5 Simulation Results 255 6.4 The Proposed Age Determination Scheme 259 6.4.1 KNN Regression 260 6.4.2 KNN-Support Vector Regression (KNN-SVR) 261 6.4.3 The Modification for Unbalanced Data 261 6.5 Experimental Results 262 6.5.1 Different Combinations in The Proposed Framework 263 6.5.2 Comparisons with Existing Algorithms 265 6.5.3 More Discussions 271 6.6 Conclusion 272 Chapter 7 Conclusion and Future work 275 7.1 Conclusion 275 7.2 Future work 276 REFERENCE 277 | |
dc.language.iso | en | |
dc.title | 整合式機器學習演算法與人臉年齡估測 | zh_TW |
dc.title | Integrated Machine Learning Algorithms for Human Age Estimation | en |
dc.type | Thesis | |
dc.date.schoolyear | 99-2 | |
dc.description.degree | 碩士 | |
dc.contributor.oralexamcommittee | 郭景明(Jing-Ming Guo),徐宏民(Winston H. Hsu),曾易聰(Yi-Chong Zeng) | |
dc.subject.keyword | 機器學習,降維,流形學習,距離度量學習,人類年齡估測,回歸, | zh_TW |
dc.subject.keyword | machine learning,dimensionality reduction,manifold learning,distance metric learning,human age estimation,regression, | en |
dc.relation.page | 298 | |
dc.rights.note | 有償授權 | |
dc.date.accepted | 2011-08-08 | |
dc.contributor.author-college | 電機資訊學院 | zh_TW |
dc.contributor.author-dept | 電機工程學研究所 | zh_TW |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-100-1.pdf 目前未授權公開取用 | 6.48 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。