You are previewing Pattern Recognition Technologies and Applications: Recent Advances.
O'Reilly logo
Pattern Recognition Technologies and Applications: Recent Advances

Book Description

"The nature of handwriting in our society has significantly altered over the ages due to the introduction of new technologies such as computers and the World Wide Web. With increases in the amount of signature verification needs, state of the art internet and paper-based automated recognition methods are necessary.

Pattern Recognition Technologies and Applications: Recent Advances provides cutting-edge pattern recognition techniques and applications. Written by world-renowned experts in their field, this easy to understand book is a must have for those seeking explanation in topics such as on- and offline handwriting and speech recognition, signature verification, and gender classification."

Table of Contents

  1. Copyright
  2. Preface
  3. Acknowledgment
  4. I. Fusion of Segmentation Strategies for Off-Line Cursive Handwriting Recognition
    1. ABSTRACT
    2. INTRODUCTION
    3. TYPICAL HANDWRITING RECOGNITION SYSTEM
      1. Preprocessing
      2. Segmentation
      3. Feature Extraction
      4. Feature Selection
      5. Classification
    4. REVIEW OF EXISTING HANDWRITING RECOGNITION TECHNIQUES/SYSTEMS
      1. State of the Art in Cursive Word Recognition
    5. PROPOSED STRATEGIES FOR SEGMENTATION-BASED HANDWRITING RECOGNITION
    6. CONCLUSION AND FUTURE RESEARCH
    7. REFERENCES
  5. II. Elastic Matching Techniques for Handwritten Character Recognition
    1. ABSTRACT
    2. INTRODUCTION
    3. OUTLINE OF ELASTIC MATCHING
      1. Formulation of EM
    4. CLASSIFICATION OF ELASTIC MATCHING TECHNIQUES
      1. Parametric 2DW
        1. Linear DW
        2. Orthogonal 2DW
      2. Nonparametric 2DW
        1. Continuous 2DW
        2. Discrete and Unconstrained 2DW
        3. Discrete and Constrained 2DW
      3. Hybrid between Parametric 2DW and Nonparametric 2DW
    5. RELATED TOPICS
      1. Comparison with Shape Normalization
      2. Reference Patterns for EM
      3. Pixel Feature
      4. Category-Dependent Deformation Tendency
    6. EM TECHNIQUES BASED ON 1D-2D MAPPING
    7. FUTURE TASKS
      1. Reduction of Computational Complexity
      2. Design of 2DW by Category-Dependent Characteristics
      3. Multistep Deformation Compensation
      4. Feature Extraction
      5. EM for Handwritten Word Recognition
      6. Utilization of Optimized 2DW
      7. EM for Camera-Based Character Recognition
      8. EM for Kernel Machines
    8. CONCLUSION
    9. ACKNOWLEDGMENT
    10. REFERENCES
  6. III. State of the Art in Off-Line Signature Verification
    1. ABSTRACT
    2. INTRODUCTION
    3. SIGNATURE AND FORGERY TYPES
    4. FEATURE EXTRACTION TECHNIQUES
      1. Representations of a Signature
      2. Geometrical Features
      3. Statistical Features
      4. Similarity Features
      5. Fixed Zoning
      6. Signal-Dependent Zoning
      7. Pseudo-Dynamic Features
      8. Discussion
    5. VERIFICATION STRATEGIES AND EXPERIMENTAL RESULTS
      1. Performance Evaluation Measures
      2. Distance Classifiers
      3. Artificial Neural Networks
      4. Hidden Markov Models
      5. Dynamic Time Warping
      6. Support Vector Machines
      7. Structural Techniques
      8. Discussion
    6. DEALING WITH A LIMITED AMOUNT OF DATA
    7. CONCLUSION
    8. REFERENCES
    9. ENDNOTE
  7. IV. An Automatic Off-Line Signature Verification and Forgery Detection System
    1. ABSTRACT
    2. INTRODUCTION
    3. HANDWRITTEN SIGNATURES
    4. HANDWRITTEN SIGNATURE FEATURES
      1. Types of Features
      2. General Overview of Signature Features
    5. DATA ACQUISITION
    6. PREPROCESSING
      1. Binarization
      2. Slant Normalization
      3. Skeletonization
      4. Smoothing
        1. Endpoint Smoothing
      5. Size Normalization
    7. FEATURE EXTRACTION
      1. Grid-Based Approaches
      2. Signature Grid Method
      3. Signature Grid Features
    8. VERIFICATION SYSTEM
      1. System Design
        1. Model Formulation
      2. Implementation
        1. Case 1: TS Model with Consequent Coefficients Fixed
        2. Case 2. TS Model with Adaptive Consequent Coefficients
      3. Experiments
    9. CONCLUSION
    10. REFERENCES
  8. V. Introduction to Speech Recognition
    1. ABSTRACT
    2. INTRODUCTION
    3. STATE-OF-THE-ART
    4. FUNDAMENTALS OF VOICE RECOGNITION
    5. GENERAL FRAMEWORK FOR SPEECH RECOGNITION
      1. Prepare the Speech Corpus to Use in the Research
      2. Requirements to Obtain a Good Quality Corpus
      3. Preprocessing Applied to the Corpus Files
        1. Energy Analysis
        2. Analysis of Zero Crossing
      4. Feature Extraction Algorithms
      5. System Training
      6. ASR Testing
    6. CONCLUSION
    7. REFERENCES
  9. VI. Seeking Patterns in the Forensic Analysis of Handwriting and Speech
    1. ABSTRACT
    2. INTRODUCTION
    3. BACKGROUND
    4. FORENSIC FEATURES OF HANDWRITING
      1. Introduction and Problem Statement
      2. Analysis of Forensic Document Features
        1. Choice of Characters for Feature Extraction
        2. Choice of Features to Study
      3. Extraction of Features
        1. Initial Image Preprocessing
        2. Extraction of Skeleton-Independent Features
        3. Extraction of Angular Features
        4. Extraction of Ascender and Descender Features
        5. Extraction of Other Features
      4. Evaluation of the Features
        1. Features of "d," "y," "f," and "th"
        2. Features of "th" with Vector Skeletonization
    5. CONCLUSION
    6. FORENSIC STUDY OF SPEECH FEATURES
      1. Speaker Verification in Forensic Analysis
      2. Structure of a Speaker Verification System
      3. Features for Speaker Verification
        1. Frame-Level Analysis
        2. Mel Frequency Cepstral Coefficients (MFCC)
        3. Mel Linear Spectrum Frequencies (MLSF)
        4. Residual Phase
        5. Hurst Parameter Features (pH)
        6. Features Based on Fractional Fourier Transform
      4. Techniques for Speaker Recognition
      5. Experiments and Results
        1. Mel Frequency Cepstral Coefficients (MFCC)
        2. Mel Linear Spectrum Frequencies (MLSF)
        3. Hurst Parameter Related Features (pH)
        4. Linear Prediction Residual Phase
        5. Features Based on Fractional Fourier Transform
        6. Combination of Features
      6. Summary of Studied Features
    7. CONCLUSION
    8. REFERENCES
  10. VII. Image Pattern Recognition-Based Morphological Structure and Applications
    1. ABSTRACT
    2. INTRODUCTION
    3. SMOOTH FOLLOWING
      1. Contour Smooth Following with Difference Codes
      2. The Smoothing Following Based on Removing Spurious Point Groups
      3. Skeleton and Its Smoothing
      4. Summary
    4. LINEARIZATION OF CONTOURS-BASED DIFFERENCE CHAIN CODES
      1. Definition of Linearizing Line Based on Difference Codes
    5. STRUCTURAL POINTS OF BINARY IMAGE CONTOURS
      1. Morphological Structural Point of Contours
      2. Experiment Results
    6. ANALYSIS AND RECOGNITION OF BROKEN HANDWRITTEN DIGITS
      1. Preprocessing Algorithms of Broken Digits
        1. Filling and Constrained Dilation
      2. Preprocessing
      3. Broken Points of Broken Digits
        1. Preselection of Broken Points
        2. Correction of Preselected Broken Points
      4. Reconstruction and Recognition of Internally Broken Handwritten Digits
        1. Case 1
        2. Case 2
      5. Experimental Results of Reconstruction
      6. Recognition of Handwritten Digits Based on Reconstruction and ONNC
      7. Recognition of Handwritten Digits Based on Morphological Structures
      8. Summary
    7. DYNAMIC ANALYSIS AND RECOGNITION OF HIGH CONTENT CELL-CYCLE SCREENING BASED ON MORPHOLOGICAL STRUCTURES
      1. The Morphological Structure of Various Cell Phases
        1. Binarization
        2. Smooth Following and Linearization
        3. The Series of Structural Points of Cell Images
      2. Separation and Reconstruction of Touching Cells
        1. Morphology Structures of Touching Cell Images
        2. Separation Points of Touching Cell Images
        3. Reconstruction of Touching Cell Images
      3. Dynamic Analysis and Recognition of Different Cell Phases
        1. Morphological Model 1: Ellipse shapes e(5,1,2,6) and e(6,2,1,5)
        2. Morphological Model 2: Ellipse shapes e(7,3,0,4) and e(0,4,3,7)
        3. Morphological Model 3: Ellipse shapes e(6,2,7,3) and e(7,3,2,6)
        4. Morphological Model 4: Ellipse shapes e(5,1,0,4) and e(0,4,5,1)
        5. Morphological Model 5: Barbell Shapes
        6. Cell Phases
      4. Summary
    8. CONCLUSION
    9. REFERENCES
  11. VIII. Robust Face Recognition Technique for a Real-Time Embedded Face Recognition System
    1. ABSTRACT
    2. INTRODUCTION
    3. BACKGROUND OF FACE RECOGNITION
      1. Illumination-Insensitive Face Recognition
      2. Expression Invariant Face Recognition
      3. Background of Pose-Insensitive Face Recognition
    4. HEAD POSE COMPENSATION
      1. Facial Feature Interpretation
        1. Statistical Models of Shape and Appearance
        2. Combination of Statistical Shape and Appearance Models
        3. Interpreting Images Using Active Shape and Appearance Models
      2. Combination of Cascade Face Detector with Active Appearance Model Search
      3. Pose Estimation
      4. Frontal View Synthesis
    5. RESULTS FROM EXPERIMENTS
      1. Training of Frontal Face Models
      2. Searching Results of Active Shape Models (ASMs) and Active Appearance Models (AAMs)
      3. Training of Correlation Model
      4. Synthesis Results from a Frontal Face Image
      5. High Pose Angle Face Recognition Results
    6. REAL-TIME AUTOMATED FACE RECOGNITION SYSTEMS
      1. Related Work
      2. VLSI-Based Face Detector
      3. Face Detection Implemented on Configurable Platforms
      4. System Level-Based Design Methodology
      5. DSP-Based Implementations
      6. Future Directions
      7. Optimization Using Custom Instructions
      8. Custom Instruction Design Flow
      9. Design Considerations
    7. CONCLUSION AND FUTURE WORK
    8. ACKNOWLEDGMENT
    9. REFERENCES
    10. ENDNOTE
  12. IX. Occlusion Sequence Mining for Activity Discovery from Surveillance Videos
    1. ABSTRACT
    2. INTRODUCTION
    3. FOREGROUND EXTRACTION
    4. TRACKING MULTIPLE OBJECTS
      1. Object Representation and Localization
      2. Identifying Occlusion Primitives
    5. ACTIVITY DISCOVERY
      1. Incremental Transition Sequence Learning
      2. Unsupervised Interaction Learning
    6. RESULTS
    7. CONCLUSION
    8. REFERENCES
  13. X. Human Detection in Static Images
    1. ABSTRACT
    2. INTRODUCTION
    3. BACKGROUND
    4. FEATURE EXTRACTION
      1. PCA Coefficients
      2. Haar-Like Features
      3. Histogram of Oriented Gradients
    5. CLASSIFICATION METHODS
      1. Support Vector Machine
      2. Cascaded AdaBoost
    6. FUSION OF MULTIPLE DETECTIONS
    7. FAST HUMAN DETECTION BY BOOSTING HOG FEATURES
      1. The Framework
      2. The HOG Feature Pool
    8. EXPERIMENTS
      1. Comparisons of Different Feature and Classifier
      2. Comparisons of Three Systems
    9. FUTURE TRENDS
    10. CONCLUSION
    11. ACKNOWLEDGMENT
    12. REFERENCE
  14. XI. A Brain-Inspired Visual Pattern Recognition Architecture and Its Applications
    1. ABSTRACT
    2. INTRODUCTION
    3. THE BRAIN-INSPIRED NEURAL NETWORK ARCHITECTURE
      1. Visual Feature Extraction
      2. Network Connection Schemes
      3. Pattern Classification Stage
      4. Shunting Neuron Model
      5. Network Training Process
    4. APPLICATIONS
      1. Visual Documents Analysis
      2. Texture Segmentation
      3. Automatic Face Detection
        1. Face Classifier
        2. Face Localization
        3. Performance of the Face Detection System
      4. Gender Recognition
    5. CONCLUSION
    6. ACKNOWLEDGMENT
    7. REFERENCES
    8. ENDNOTES
  15. XII. Significance of Logic Synthesis in FPGA-Based Design of Image and Signal Processing Systems
    1. ABSTRACT
    2. INTRODUCTION
    3. BASIC THEORY
      1. Cube Representation of Boolean Functions
      2. Representation and Analysis of Boolean Functions with Blankets
      3. Example 1: Blanket-Based Representation of Boolean Functions
      4. Serial Decomposition
      5. Theorem 1: Existence of Serial Decomposition (Brzozowski & Łuba, 2003)
      6. Example 2
      7. Parallel Decomposition
      8. Example 3
      9. Balanced Functional Decomposition
      10. Example 4
    4. DIGITAL FILTERS
    5. DISTRIBUTED ARITHMETIC METHOD
    6. RESULTS
    7. CONCLUSION
    8. ACKNOWLEDGMENT
    9. REFERENCES
  16. XIII. A Novel Support Vector Machine with Class-Dependent Features for Biomedical Data
    1. ABSTRACT
    2. INTRODUCTION
    3. BACKGROUND
    4. METHODOLOGY
      1. Feature Importance Ranking Measures
      2. RELIEF Ranking Measure
      3. Class Separability Measure
      4. The Classifier: The Support Vector Machine
      5. Our Class-Dependent Feature Selection Approach
      6. The Constructed Novel SVM Classifier for Class-Dependent Features
    5. EXPERIMENTS AND ANALYSIS
      1. Experimental Data
      2. Software Preparations
      3. Experiment Results and Analysis
    6. CONCLUSION
    7. REFERENCES
  17. XIV. A Unified Approach to Support Vector Machines
    1. ABSTRACT
    2. INTRODUCTION
    3. SUPPORT VECTOR CLASSIFIERS
      1. The Separable Case
      2. Example: XOR Gate
      3. The Inseparable Case
      4. Support Vector Machines and the Risk Bound
    4. ONE-CLASS SUPPORT VECTOR CLASSIFIERS
    5. SUPPORT VECTOR REGRESSORS
      1. Example: XOR Gate Revisited
    6. UNIFICATION AND DUALITY
      1. Unification
      2. The Dual Formulation
      3. Properties of the Dual
      4. Special Cases
        1. The C-SVC
        2. The One-Class C-SVC
        3. The ε-SVR
    7. A MECHANICAL ANALOGY
    8. GENERAL COST FUNCTIONS IN THE UNIFIED SVM
    9. A PRACTICAL EXAMPLE: SPAM DETECTION
    10. CONCLUSION
    11. REFERENCES
    12. ENDNOTES
  18. XV. Cluster Ensemble and Multi-Objective Clustering Methods
    1. ABSTRACT
    2. INTRODUCTION
    3. BACKGROUND
    4. MULTI-OBJECTIVE CLUSTERING ENSEMBLE
    5. APPLICATION
    6. FUTURE TRENDS
    7. CONCLUSION
    8. REFERENCES
  19. XVI. Implementing Negative Correlation Learning in Evolutionary Ensembles with Suitable Speciation Techniques1
    1. ABSTRACT
    2. INTRODUCTION
    3. DIVERSITY AND MOTIVATION FOR ENSEMBLES
      1. Importance of Diversity
      2. Ensembles of Independent Estimators
      3. The Ambiguity Decomposition
      4. Bias-Variance Dilemma
      5. Bias, Variance, and Covariance
      6. Diversity Creation Taxonomy
    4. NEGATIVE CORRELATION LEARNING
      1. Rosen's Decorrelation Penalty Term
      2. Original NCL Penalty Function
      3. Amended Derivation of NCL Penalty Function
      4. Additional NCL Developments
    5. SINGLE-OBJECTIVE EVOLUTION AND NCL
      1. EENCL Algorithm
        1. Algorithm Description
        2. Experiments with Local Search in EENCL
        3. Results and Discussion
      2. INCL Algorithm
        1. Algorithm Description
      3. Experiments with Speciation Techniques in Evolutionary Ensembles
      4. Results and Discussion
      5. Comparison of INCL and EENCL to Alternative Ensemble Techniques
      6. Error Rates
      7. Bias, Variance, and Covariance Analysis
      8. Discussion
    6. CONCLUSION
    7. REFERENCES
    8. ENDNOTE
  20. XVII. A Recurrent Probabilistic Neural Network for EMG Pattern Recognition
    1. ABSTRACT
    2. INTRODUCTION
    3. BACKGROUND
    4. A RECURRENT PROBABILISTIC NEURAL NETWORK
      1. HMM-Based Dynamic Probabilistic Model
      2. Network Architecture
      3. A Maximum Likelihood Training Algorithm
    5. EMG PATTERN RECOGNITION USING R-LLGMN
      1. Experimental Conditions
      2. Pattern Recognition of Filtered EMG Signals
      3. Pattern Recognition of Raw EMG Signals
    6. DISCUSSIONS
    7. SUMMARY
    8. REFERENCES
  21. Compilation of References
  22. About the Contributors