Contents Chapter 1 Recent Advances in Pattern Classification Marek R. Ogiela, Lakhmi C Jain 1 New Directions in Pattern Classification References Chapter 2 Neural Networks for Handwriting Recognition Marcus Liwicki. Alex graves. horst bunke 1 Introduction 1.1 State-of-the-Art 1. 2 Contribution 2 Data processing 2.1 General Processing Steps 2.2 Our Online System 2.3 Our Offline System 3 Neural Network Based Recognition 3.1 Recurrent Neural Networks(RNNs) 3.2 Long Short-Term Memory (LSTM) 3.3 Bidirectional Recurrent Neural Networks 3.4 Connectionist Temporal Classification(CTC) 3.5 Multidimensional Recurrent Neural Networks 3.6 Hierarchical Subsampling Recurrent Neural Networks Experiments…… 4.1 Comparison with HMMs on the IAM Databases.... 4.2 Recognition Performance of MdlStM on Contest'Data 5 Conclusion References Chapter 3 Moving Object Detection from Mobile Platforms Using Stereo Data Angel D Sappa, David Geronimo, Fadi Dornaika, Mohammad Rouhani Antonio M. Lopez 1 Introduction 2 Related Work
Contents Chapter 1 Recent Advances in Pattern Classification..........................................................1 Marek R. Ogiela, Lakhmi C. Jain 1 New Directions in Pattern Classification......................................................1 References .........................................................................................................4 Chapter 2 Neural Networks for Handwriting Recognition..................................................5 Marcus Liwicki, Alex Graves, Horst Bunke 1 Introduction ..................................................................................................5 1.1 State-of-the-Art.....................................................................................6 1.2 Contribution..........................................................................................7 2 Data Processing ............................................................................................8 2.1 General Processing Steps......................................................................9 2.2 Our Online System .............................................................................10 2.3 Our Offline System.............................................................................12 3 Neural Network Based Recognition ...........................................................12 3.1 Recurrent Neural Networks (RNNs)...................................................12 3.2 Long Short-Term Memory (LSTM) ...................................................13 3.3 Bidirectional Recurrent Neural Networks...........................................16 3.4 Connectionist Temporal Classification (CTC) ...................................16 3.5 Multidimensional Recurrent Neural Networks ...................................17 3.6 Hierarchical Subsampling Recurrent Neural Networks......................18 4 Experiments ................................................................................................18 4.1 Comparison with HMMs on the IAM Databases................................18 4.2 Recognition Performance of MDLSTM on Contest’ Data .................20 5 Conclusion ..................................................................................................21 References .......................................................................................................21 Chapter 3 Moving Object Detection from Mobile Platforms Using Stereo Data Registration......................................................................................................... 25 Angel D. Sappa, David Gerónimo, Fadi Dornaika, Mohammad Rouhani, Antonio M. López 1 Introduction ............................................................................................... 25 2 Related Work............................................................................................. 26
Contents 3 Proposed Approach 3.1 System Setup 3.2 Feature Detection and Trackin 3.3 Robust registration 3.4 Frame Subtraction Experimental Results 32 References Chapter 4 Pattern Classifications in Cognitive Informatics Lidia Ogiela 1 Introduction .39 2 Semantic Analysis Stages 3 Semantic Analysis vs. Cognitive Informatics 4 Example of a Cognitive UBIAS System References Optimal Differential Filter on Hexagonal Lattice. Suguru Saito, Masayuki Nakajiama, Tetsuo Shima 2 Preliminaries 60 3 Least Inconsistent Image .60 4 Point Spread Function 5 Condition for gradient Filter 67 6 Numerical Optimization.. 68 7 Theoretical Evaluation 7. 1 Signal-to-Noise Ratio 8 Experimental Evaluation. 8.1 Construction of Artificial Images 8.2 Detection of Gradient Intensity and Orientation 8.3 Overington's Method of Orientation Detection 8.4 Relationship between Derived Filter and Staunton Filter 8.5 Experiment and Results 81 10 Summary. References
VIII Contents 3 Proposed Approach.................................................................................... 28 3.1 System Setup ..................................................................................... 29 3.2 Feature Detection and Tracking......................................................... 29 3.3 Robust Registration ........................................................................... 31 3.4 Frame Subtraction.............................................................................. 31 4 Experimental Results ................................................................................. 34 5 Conclusions ............................................................................................... 35 References ...................................................................................................... 35 Chapter 4 Pattern Classifications in Cognitive Informatics ............................................ 39 Lidia Ogiela 1 Introduction ............................................................................................... 39 2 Semantic Analysis Stages .......................................................................... 40 3 Semantic Analysis vs. Cognitive Informatics ............................................ 43 4 Example of a Cognitive UBIAS System.................................................... 45 5 Conclusions ............................................................................................... 52 References ...................................................................................................... 52 Chapter 5 Optimal Differential Filter on Hexagonal Lattice............................................ 59 Suguru Saito, Masayuki Nakajiama, Tetsuo Shima 1 Introduction ............................................................................................... 59 2 Preliminaries.............................................................................................. 60 3 Least Inconsistent Image ........................................................................... 60 4 Point Spread Function................................................................................ 64 5 Condition for Gradient Filter ..................................................................... 67 6 Numerical Optimization ............................................................................ 68 7 Theoretical Evaluation............................................................................... 70 7.1 Signal-to-Noise Ratio ........................................................................ 70 7.2 Localization ....................................................................................... 73 8 Experimental Evaluation............................................................................ 74 8.1 Construction of Artificial Images ...................................................... 74 8.2 Detection of Gradient Intensity and Orientation................................ 76 8.3 Overington's Method of Orientation Detection.................................. 76 8.4 Relationship between Derived Filter and Staunton Filter .................. 78 8.5 Experiment and Results ..................................................................... 80 9 Discussion.................................................................................................. 81 10 Summary.................................................................................................. 83 References ...................................................................................................... 86
Contents Chapter 6 Graph Image Language Techniques Supporting Advanced Classification and Cognitive Interpretation of CT Coronary Vessel Visualizations 1 Introduction 2 The Classification problem 3 Stages in the Analysis of CT Images under a Structural Approach Utilising Graph Techniques 4 Parsing Languages Generated by Graph Grammars. 5 Picture Grammars in Classification and Semantic Interpretation of 3D Coronary Vessels Visualisations 5.1 Characteristics of the lmage data 5.2 Preliminary Analysis of 3D Coronary Vascularisation 5.3 Graph-Based Linguistic Formalisms in Spatial Modelling of Coronary Vessels 5.4 Detecting Lesions and Constructing the Syntactic Analyser 6 Conclusions Concerning the Advanced Classification and Cognitive 104 5.5 Selected Results Interpretation of CT Coronary Vessel Visualizations 10 Chapter 7 A Graph Matching Approach to Symmetry Detection and analysis Michael chertok and yosi keller 1 Introduction 113 2 Symmetries and Their Properties 115 2.1 Rotational Symmetry 2.2 Reflectional Symmetry 116 2.3 Interrelations between Rotational and Reflectional Symmetries. 117 2.4 Discussion 3 Previous Work 117 3.1 Previous Work in Symmetry Detection and Analysis 118 3.2 Local features 3.3 Spectral Matching of Sets of Points in R Spectral Symmetry Analysis 4.1 Spectral Symmetry Analysis of Sets in R 4. 1. 1 Perfect Symmetry and Spectral De 4.2 Spectral Symmetry Analysis of Images 4.2. 1 Image Representation by Local Features 125 4.2.2 Symmetry Categorization and Pruning 4.2.3 Computing the Geometrical Properties of the Symmetry..126 5 Experimental Results 127 5.1 Symmetry Analysis of Images... 5.2 Statistical Accuracy Analysis 134
Contents IX Chapter 6 Graph Image Language Techniques Supporting Advanced Classification and Cognitive Interpretation of CT Coronary Vessel Visualizations ............ 89 Mirosław Trzupek 1 Introduction ............................................................................................... 89 2 The Classification Problem........................................................................ 92 3 Stages in the Analysis of CT Images under a Structural Approach Utilising Graph Techniques ....................................................................... 93 4 Parsing Languages Generated by Graph Grammars .................................. 95 5 Picture Grammars in Classification and Semantic Interpretation of 3D Coronary Vessels Visualisations ............................................................... 96 5.1 Characteristics of the Image Data ...................................................... 96 5.2 Preliminary Analysis of 3D Coronary Vascularisation Reconstructions.................................................................................. 96 5.3 Graph-Based Linguistic Formalisms in Spatial Modelling of Coronary Vessels ............................................................................... 98 5.4 Detecting Lesions and Constructing the Syntactic Analyser ........... 103 5.5 Selected Results ............................................................................... 104 6 Conclusions Concerning the Advanced Classification and Cognitive Interpretation of CT Coronary Vessel Visualizations............................. 108 References .................................................................................................... 110 Chapter 7 A Graph Matching Approach to Symmetry Detection and Analysis............113 Michael Chertok and Yosi Keller 1 Introduction ..............................................................................................113 2 Symmetries and Their Properties..............................................................115 2.1 Rotational Symmetry ........................................................................115 2.2 Reflectional Symmetry .....................................................................116 2.3 Interrelations between Rotational and Reflectional Symmetries ......117 2.4 Discussion.........................................................................................117 3 Previous Work ..........................................................................................117 3.1 Previous Work in Symmetry Detection and Analysis.......................118 3.2 Local Features...................................................................................121 3.3 Spectral Matching of Sets of Points in Rn .........................................122 4 Spectral Symmetry Analysis.....................................................................123 4.1 Spectral Symmetry Analysis of Sets in Rn .......................................123 4.1.1 Perfect Symmetry and Spectral Degeneracy..........................124 4.2 Spectral Symmetry Analysis of Images ............................................124 4.2.1 Image Representation by Local Features ...............................125 4.2.2 Symmetry Categorization and Pruning ..................................125 4.2.3 Computing the Geometrical Properties of the Symmetry ......126 5 Experimental Results ................................................................................127 5.1 Symmetry Analysis of Images ..........................................................128 5.2 Statistical Accuracy Analysis ...........................................................134
Contents 5.3 Analysis of Three-Dimensional Symmetry. 5.5 Additional Results 140 6 Conclusions 140 References Chapter 8 Pattern Classification Methods for Analysis and visualization of Brain Perfusion CT Maps 145 Hach 1 Introduction 2 Interpretation of Perfusion Maps-Long and Short Time Prognosis.. 148 3 Image Processing and Abnormality Detection 4 Image Registration 153 4.1 Affine Registration 4.2 FFD Registration 4.3 Thirion's Demons algorithm 154 4.4 Comparison of Registration Algorithms 5 Classification of Detected Abnormalities 6 System Validation and Results 7 Data Visualization- Augmented Reality Environment 162 7.1 Augmented Reality Environment 7.2 Real Time Rendering of 3D Data 7.3 Augmented Desktop-System Performance Test 164 8 Summary. References 168 Chapter 9 Inference of Co-occurring Classes: Multi-class and Multi-label Classification 171 Tal Sobol-Shikle 1 Introduction 2 Applications 3 The Classification Process 173 4 Data and Annotation .175 5 Classification Approaches 5.1 Binary Classification 5.2 Multi-class Classification 177 5.3 Multi-label Classification 6 Multi-class Classification 6. 1 Multiple Binary Classifiers 6.1.1 One-Against-All Classification 6.1.2 One-Against-One(Pair-Wise)Classification 6.1.3 Combining Binary Classifiers
X Contents 5.3 Analysis of Three-Dimensional Symmetry.......................................136 5.4 Implementation Issues ......................................................................137 5.5 Additional Results ............................................................................140 6 Conclusions ..............................................................................................140 References .....................................................................................................142 Chapter 8 Pattern Classification Methods for Analysis and Visualization of Brain Perfusion CT Maps............................................................................................145 Tomasz Hachaj 1 Introduction ..............................................................................................145 2 Interpretation of Perfusion Maps – Long and Short Time Prognosis........148 3 Image Processing and Abnormality Detection..........................................150 4 Image Registration....................................................................................153 4.1 Affine Registration ...........................................................................154 4.2 FFD Registration ..............................................................................154 4.3 Thirion’s Demons Algorithm............................................................154 4.4 Comparison of Registration Algorithms ...........................................155 5 Classification of Detected Abnormalities .................................................158 6 System Validation and Results .................................................................160 7 Data Visualization – Augmented Reality Environment............................162 7.1 Augmented Reality Environment .....................................................163 7.2 Real Time Rendering of 3D Data .....................................................164 7.3 Augmented Desktop - System Performance Test .............................164 8 Summary...................................................................................................167 References .....................................................................................................168 Chapter 9 Inference of Co-occurring Classes: Multi-class and Multi-label Classification ......................................................................................................171 Tal Sobol-Shikler 1 Introduction ..............................................................................................171 2 Applications..............................................................................................172 3 The Classification Process ........................................................................173 4 Data and Annotation .................................................................................175 5 Classification Approaches ........................................................................177 5.1 Binary Classification.........................................................................177 5.2 Multi-class Classification .................................................................177 5.3 Multi-label Classification .................................................................178 6 Multi-class Classification .........................................................................179 6.1 Multiple Binary Classifiers...............................................................180 6.1.1 One-Against-All Classification..............................................180 6.1.2 One-Against-One (Pair-Wise) Classification.........................180 6.1.3 Combining Binary Classifiers................................................181
Contents 6.2 Direct Multi-class Classification 6.3 Associative Classification 7 Multi-label Classification 7.1 Semi-supervised(Annotation) Methods 8 Inference of Co-occurring Affective States from Non-verbal Speech 9 Summary References Author Index
Contents XI 6.2 Direct Multi-class Classification.......................................................181 6.3 Associative Classification.................................................................182 7 Multi-label Classification .........................................................................182 7.1 Semi-supervised (Annotation) Methods ...........................................186 8 Inference of Co-occurring Affective States from Non-verbal Speech......186 9 Summary...................................................................................................193 References .....................................................................................................193 Author Index......................................................................................................199