Biometric Recognition: 15th Chinese Conference, CCBR 2021, Shanghai, China, September 10–12, 2021, Proceedings (Image Processing, Computer Vision, Pattern Recognition, and Graphics)

Preface
Organization
Contents
Multi-modal Biometrics and Emerging Biometrics
A Novel Dual-Modal Biometric Recognition Method Based on Weighted Joint Sparse Representation Classifaction
	1 Introduction
	2 Multimodal Joint Sparse Representation Classification
	3 Proposed Method
	4 Experiments
		4.1 Datasets
		4.2 The Experiments for Selecting Parameter Values
		4.3 Comparison Experiments
	5 Conclusion
	References
Personal Identification with Exploiting Competitive Tasks in EEG Signals
	1 Introduction
	2 Related Work
		2.1 EEG Based Personal Identification
		2.2 Multi-task Adversarial Learning
	3 Method
		3.1 LFCC
		3.2 Adversarial Multi-task Learning
	4 Experiment
		4.1 Dataset
		4.2 Implementation Details
		4.3 Ablation Study
		4.4 Comparison with Other Methods
	5 Conclusion
	References
A Systematical Solution for Face De-identification
	1 Introduction
	2 Background and Related Work
	3 Method
		3.1 Attribute Disentanglement and Generative Network
		3.2 Adversarial Vector Mapping Network
	4 Experiment
		4.1 De-ID by Adversarial Attacks
		4.2 De-ID by Face Swapping
	5 Conclusion
	References
Skeleton-Based Action Recognition with Improved Graph Convolution Network
	1 Introduction
	2 Method
		2.1 Approach Overview
		2.2 Improved GCN
		2.3 Cluster Training
	3 Experiment
		3.1 Datasets
		3.2 Implementation Details
		3.3 Ablation Learning
		3.4 Compared with the State-of-the-Art
	4 Conclusion
	References
End-To-End Finger Trimodal Features Fusion and Recognition Model Based on CNN
	1 Introduction
	2 Integrated Acquisition System
	3 Preliminaries
		3.1 Asymmetric Convolution Kernel
		3.2 Multi-channel Convolution
	4 Model Architecture
	5 Experiments and Analysis
		5.1 Experimental Dataset
		5.2 Experimental Results and Analysis
	6 Conclusion
	References
Mouse Dynamics Based Bot Detection Using Sequence Learning
	1 Introduction
	2 Related Works and Background
		2.1 Mouse Dynamics
		2.2 Preliminary Knowledge
	3 Framework of Our Bot Detection
	4 Data Collection
		4.1 Mouse-Operation Task Design
		4.2 Human Data Collection
		4.3 Web Bot Data Collection
	5 Experiments and Analysis
		5.1 Models Implementation
		5.2 Evaluation Metrics
		5.3 Training and Testing Procedure
		5.4 Comparison Among Different Time Series Representations
		5.5 Comparison with Baselines in Bot Detection
	6 Conclusion
	References
A New Age-Groups Classifying Method for Irrawaddy Dolphin
	1 Introduction
	2 Proposed Method
		2.1 Methodology
	3 Experiments
		3.1 Data Collection and Feature Selection
		3.2 Experiments and Results
	4 Conclusions and Discussions
	References
Auricular Point Localization Oriented Region Segmentation for Human Ear
	1 Introduction
	2 Improved YOLACT Model with Selecting Mask Module
		2.1 Protonet Module and Target Detection Module
		2.2 Selecting Mask Module
		2.3 Loss Function
	3 Ear Region Segmentation Based on Improved YOLACT
	4 Conclusion
	References
Portrait Thangka Image Retrieval via Figure Re-identification
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Challenges in Portrait Thangka Image Retrieval
		3.2 Re-identification Based Portrait Thangka Image Retrieval
	4 Experiments
	5 Conclusions and Future Work
	References
To See Facial Expressions Through Occlusions via Adversarial Disentangled Features Learning with 3D Supervision
	1 Introduction
	2 Proposed Method
		2.1 Overview
		2.2 3D Supervised Feature Disentanglement
		2.3 Adversarial Training Against Occlusion Noise
		2.4 The Learning Process
	3 Experiments
		3.1 Datasets and Implementation Details
		3.2 Results
	4 Conclusion
	References
Automatically Distinguishing Adult from Young Giant Pandas Based on Their Call
	1 Introduction
	2 Related Work
	3 Dataset
		3.1 Data Acquisition
		3.2 Data Preprocessing
	4 Method
		4.1 Acoustic Feature
		4.2 Input Feature
		4.3 Network Structure
	5 Experiments
		5.1 The Effectiveness of Augmentation
		5.2 Compare the Effectiveness of MFCC and Pre-emphasis
		5.3 Compare the Effectiveness of MFCC and IMFCC
		5.4 Discussion on the Results
	6 Conclusion
	References
Alzheimer\'s Disease Prediction via the Association of Single Nucleotide Polymorphism with Brain Regions
	1 Introduction
	2 Material and Methodology
		2.1 Data Acquisition and Preprocessing
		2.2 Construction of Fusion Features
		2.3 Model Fusion for Prediction
	3 Experimental Results
		3.1 Model Performance and Comparison
		3.2 Abnormal Brain Regions Prediction
		3.3 Predicting Risk Genes
	4 Discussion
		4.1 Comparison with Existing Studies
		4.2 Abnormal Brain Regions and Pathogenic Genes
		4.3 Limitations and Future Work
	5 Conclusion
	References
A Deep Attention Transformer Network for Pain Estimation with Facial Expression Video
	1 Introduction
	2 Proposed Method
	3 Experimental Results
	4 Conclusion
	References
Cognitive Analysis of EEG Signals Induced by Visual Stimulation of Facial Emotion
	1 Introduction
	2 Method
		2.1 Feature Extraction
		2.2 Channel Selection
		2.3 Decision Fusion
	3 Experiments
		3.1 Dataset
		3.2 Data Preprocessing
		3.3 Stimulation Recognition
		3.4 Depression Recognition
	4 Conclusion
	References
3D Context-Aware PIFu for Clothed Human Reconstruction
	1 Introduction
	2 Related Work
	3 Method
		3.1 Pixel-Aligned Implicit Function
		3.2 3D Context-Aware PIFu
		3.3 Multi-view Implicit Rendering
		3.4 Loss Function
	4 Experiments
		4.1 Comparison to State-of-the-Art Methods
	5 Ablation Study
		5.1 Performance Analysis
	6 Discussion and Conclusion
	References
Facial Expression Synthesis with Synchronous Editing of Face Organs
	1 Introduction
	2 Proposed Method
		2.1 Global Spatial Interaction Mechanism
		2.2 Spectrum Restriction Loss
	3 Experiment
		3.1 Experiment Settings
		3.2 Comparisons with State-of-Art Methods
		3.3 Ablation Study
		3.4 Qualitative Results on CASIA-Oulu
	4 Conclusion
	References
Multi-lingual Hybrid Handwritten Signature Recognition Based on Deep Residual Attention Network
	1 Introduction
	2 Related Theories
		2.1 Transfer Learning
		2.2 SE-ResNet
		2.3 Focal Loss
	3 Database and Pre-processing
		3.1 Signature Database
		3.2 Pre-processing
	4 Training and Classification
		4.1 GPDS Signature Recognition
		4.2 Multi-lingual Signature Recognition
	5 Conclusion
	References
Traumatic Brain Injury Images Classification Method Based on Deep Learning
	1 Introduction
	2 Methods
		2.1 The SE Module
		2.2 The PCR Module
		2.3 Network Training
	3 Experiments and Results
	4 Conclusions
	References
Palatal Rugae Recognition via 2D Fractional Fourier Transform
	1 Introduction
	2 The Proposed Palatal Rugae Recognition System
	3 Experimental Results and Analysis
	4 Conclusions and Prospects
	References
Hand Biometrics
Fusion of Partition Local Binary Patterns and Convolutional Neural Networks for Dorsal Hand Vein Recognition
	1 Introduction
	2 Related Works
		2.1 Data Argument
		2.2 PLBP
		2.3 CNNs
	3 Fusion of PLBP and CNNs
		3.1 SF
		3.2 DF
		3.3 FF
	4 Experiments and Results
		4.1 Data Preparation
		4.2 Fusion of PLBP and CNNs
	5 Conclusions
	References
Pose-Specific 3D Fingerprint Unfolding
	1 Introduction
	2 Proposed Method
		2.1 Unfolding
		2.2 3D Pose Estimation
	3 Experiment
		3.1 Database
		3.2 Matching Score
		3.3 Deformation Field
		3.4 Efficiency
	4 Conclusion
	References
Finger Vein Recognition Using a Shallow Convolutional Neural Network
	1 Introduction
	2 Related Work
	3 Proposed Method
	4 Experiments
	5 Conclusion
	References
Finger Crystal Feature Recognition Based on Graph Convolutional Network
	1 Introduction
	2 Crystal Graph Construction Method
		2.1 Node Generation
		2.2 Graph Generation and Crystal Generation
	3 Crystal Graph Classification Model
		3.1 Graph Convolutional Network
		3.2 The Improved SAGPool Model
	4 Experiments
	5 Conclusions
	References
Signatured Fingermark Recognition Based on Deep Residual Network
	1 Introduction
	2 FR-DRN Algorithm
		2.1 Extract Fingerprint Minutiae
		2.2 Processing Training Sample and Normalization
		2.3 Convolutional Neural Network Based on Residual Block
	3 Experiment
		3.1 Dataset
		3.2 Experiment Evaluation Metrics
		3.3 Experiment Results and Analysis
	4 Conclusion
	References
Dorsal Hand Vein Recognition Based on Transfer Learning with Fusion of LBP Feature
	1 Introduction
	2 Background
		2.1 Local Binary Patterns
		2.2 ResNet Network Using Transfer Learning
	3 The Proposed Method
	4 Experiments and Discussion
		4.1 Personal and Gender Recognition Based on the Original Database
		4.2 Personal and Gender Recognition Based on the Image Enhancement and Data Augmentation
	5 Conclusion
	References
An Improved Finger Vein Recognition Model with a Residual Attention Mechanism
	1 Introduction
	2 Related Works
	3 Proposed Method
		3.1 Model Architecture
		3.2 Residual Attention Block
		3.3 Attention Residual Learning
	4 Experiments
		4.1 Datasets
		4.2 Comparison Experiments
		4.3 Performance Analysis
	5 Conclusion
	References
A Lightweight CNN Using HSIC Fine-Tuning for Fingerprint Liveness Detection
	1 Introduction
	2 Proposed Methods
		2.1 Model Structure
		2.2 HSIC Fine-Tuning
	3 Experiments
		3.1 Datasets and Evaluation Metrics
		3.2 Experiment on Lightweight CNN-SPP Net
	4 Conclusions
	References
An Efficient Joint Bayesian Model with Soft Biometric Traits for Finger Vein Recognition
	1 Introduction
	2 Background
		2.1 Soft Biometric Traits
		2.2 Joint Bayesian Verification
	3 The Proposed Method
		3.1 Data Preprocessing
		3.2 Network Structure
		3.3 Joint Bayesian Recognition Loss
		3.4 Joint Bayesian Model
	4 Experimental and Results
		4.1 Comparison Between Different Inputs
		4.2 Comparison Between Different Loss Function
		4.3 Comparison with the State-of-the-Art Methods
	5 Conclusion
	References
A Novel Local Binary Operator Based on Discretization for Finger Vein Recognition
	1 Introduction
	2 Related Work
	3 Method
	4 Experiments
	5 Conclusion
	References
A Generalized Graph Features Fusion Framework for Finger Biometric Recognition
	1 Introduction
	2 Related Work
		2.1 Finger Fusion Recognition
		2.2 Graph Convolutional Neural Network
		2.3 Graph Feature Extraction and Recognition
	3 Multimodal Biometrics Fusion Framework
		3.1 Construction of Graphs
		3.2 Node Normalization
		3.3 Fusion Frameworks
	4 Experimental Results
		4.1 Unimodal Recognition
		4.2 Trimodal Fusion Recognition
	5 Conclusion
	References
A STN-Based Self-supervised Network for Dense Fingerprint Registration
	1 Introduction
	2 Related Work
		2.1 Sparse Minutiae Registration of Fingerprints
		2.2 Dense Pixel-Level Registration of Fingerprints
	3 Proposed Registration Method
		3.1 Registration Network
		3.2 Loss Function
	4 Experiments and Results
		4.1 Dataset
		4.2 Training Parameters
		4.3 Results
	5 Conclusion
	References
An Arcloss-Based and Openset-Test-Oriented Finger Vein Recognition System
	1 Introduction
	2 Methods
		2.1 Image Pre-processing
		2.2 Process and Strategy for Training and Testing of Feature Extraction Networks
	3 Results and Analysis
		3.1 Closed-Set Test
		3.2 Open-Set Test
		3.3 Mixed Open-Set Test and Closed-Set Test on SDUMLA
		3.4 Ablation Test
	4 Conclusion
	References
Different Dimension Issues in Deep Feature Space for Finger-Vein Recognition
	1 Introduction
	2 Related Work
	3 Methodology and Network Topology
		3.1 Finger-Vein ROI Extraction
		3.2 Network Topology of LFVRN
	4 Experiments and Results
		4.1 Finger-Vein Recognition Using Original Images
		4.2 Comparison of Time Efficiency
		4.3 Finger-Vein Recognition Using Enhanced Images
		4.4 Qualitative Analysis of Deep Models
	5 Conclusion
	References
Facial Biometrics
Holistic Co-occurrence Prior for High-Density Face Detection
	1 Introduction
	2 Face Co-occurrence Prior Based on Density Map
	3 Experimental Evaluation
		3.1 Dataset Preparation and Experimental Setting
		3.2 Ablation Experiments with Peer Comparision
	4 Conclusion
	References
Iris Normalization Beyond Appr-Circular Parameter Estimation
	1 Introduction
	2 Related Work
	3 Technical Details
	4 Experiments and Analysis
		4.1 Databases
		4.2 Evaluations
		4.3 Recognition Experiments
		4.4 Ablation Study
	5 Conclusion
	References
Non-segmentation and Deep-Learning Frameworks for Iris Recognition
	1 Introduction
	2 Related Work
		2.1 Iris Recognition Algorithms Based on Traditional Methods
		2.2 Iris Recognition Algorithms Based on Deep Learning
	3 Proposed Method
		3.1 ETENet
	4 Experiments and Results
		4.1 Dataset
		4.2 Experimental Environment and Parameters
		4.3 Comparison with Other Experiments
		4.4 Impact of ETENet Block
	5 Conclusion
	References
Incomplete Texture Repair of Iris Based on Generative Adversarial Networks
	1 Introduction
	2 Related Work
		2.1 Traditional Image Restoration Methods
		2.2 Deep Learning Image Restoration Methods
	3 Method of This Paper
		3.1 Network Architecture
		3.2 Evaluation Index
	4 Experiments and Results
		4.1 Description of Databases and Hardware Environment
		4.2 Experimental Program and Result Analysis
	5 Conclusion
	References
Deepfakes Detection Based on Multi Scale Fusion
	1 Introduction
	2 Relative Work
	3 Our Method
		3.1 Method Flow
		3.2 Spatial Feature Extraction
		3.3 Time Feature Extraction
	4 Experimental Results and Analysis
		4.1 Deepfakes Datasets
		4.2 Analysis of Experimental Results
		4.3 Model Performance Comparison
	5 Conclusion
	References
Balance Training for Anchor-Free Face Detection
	1 Introduction
	2 Related Work
		2.1 Scale Balance Data Augmentation
		2.2 Anchor Free Detection
	3 Proposed Methods
		3.1 Crop-Mosaic Sample Augmentation
		3.2 Density Balance Scale Assignment
		3.3 Scale Normalization Loss
	4 Experiments
		4.1 Dataset
		4.2 Implementation Details
	5 Conclusion
	References
One-Class Face Anti-spoofing Based on Attention Auto-encoder
	1 Introduction
	2 The Proposed Method
		2.1 Attention Auto-encoder
		2.2 Reconstruction Loss and Center Loss
		2.3 Spoofness Score
	3 Experiments
		3.1 Datasets
		3.2 Performance Metrics
		3.3 Experimental Settings
		3.4 Results
	4 Conclusion
	References
Full Quaternion Matrix and Random Projection for Bimodal Face Template Protection
	1 Introduction
	2 Proposed Method
		2.1 Enrollment Process
		2.2 Authentication Process
	3 Experimental Results
		3.1 Recognition Results
		3.2 Analysis of Security
		3.3 Comparison with Other Methods
	4 Conclusion
	References
Kinship Verification via Reference List Comparison
	1 Introduction
	2 Related Work
	3 Reference List Comparison Method
		3.1 Comparison of Reference Lists
		3.2 Similarity Measure
	4 Experiments
		4.1 Datasets
		4.2 Evaluation Metrics
		4.3 Experimental Settings
		4.4 Ablation Studies
		4.5 Comparison with Other Methods
	5 Conclusion
	References
Face Attribute Estimation with HMAX-GCNet Model
	1 Introduction
	2 Related Work
		2.1 Robust Object Recognition with Cortex-Like Mechanisms
		2.2 Research Status of HMAX Model
	3 Method
		3.1 Attribute Distribution Characteristics
		3.2 HMAX-GCNet Model
	4 Experiment
		4.1 Dataset
		4.2 Implementation Details
		4.3 Relationship Between Image Patches and Attributes
		4.4 Ablation Study
		4.5 Comparison with Other Methods
	5 Conclusion and Outlook
	References
Wavelet-Based Face Inpainting with Channel Relation Attention
	1 Introduction
	2 Method
		2.1 Wavelet Transfer
		2.2 Network Architecture
		2.3 Channel Relation Attention
		2.4 Wavelet Loss
	3 Experiments
		3.1 Experimental Setup
		3.2 Comparison
		3.3 Abaltion Study
	4 Conclusion
	References
Monocular 3D Target Detection Based on Cross-Modal and Mass Perceived Loss
	1 Introduction
	2 Related Works
	3 Proposed Method
		3.1 Network Frameworks
		3.2 Cross-Modal Feature Fusion Network
		3.3 3D Quality Perception Loss
	4 Experimental Results and Analysis
		4.1 The Experimental Results and Analysis
		4.2 Compare with Other Advanced Algorithms
		4.3 Visual Qualitative Analysis
	5 Conclusions
	References
Low-Quality 3D Face Recognition with Soft Thresholding
	1 Introduction
	2 Related Work
	3 Proposed Method
		3.1 Network Architecture
		3.2 Soft Thresholding Module
	4 Experiments
		4.1 Datasets
		4.2 Quantitative Evaluation
		4.3 Qualitative Evaluation
	5 Conclusion
	References
Research on Face Degraded Image Generation Algorithm for Practical Application Scenes
	1 Introduction
	2 Structure of DGAN
		2.1 Generator of DGAN
		2.2 Discriminator of DGAN
		2.3 Loss Function
	3 Experiments
		3.1 Output Image of DGAN
		3.2 Reconstruction Results of DG-DIC/DG-DICGAN
	4 Conclusion
	References
Embedding Fast Temporal Information Model to Improve Face Anti-spoofing
	1 Introduction
	2 Proposed Framework
		2.1 Dimensionality Reduction for Fast TIM
		2.2 Network for Fast TIM
	3 Experiments
		3.1 Database and Evaluation Metrics
		3.2 Implementation Details
		3.3 Experimental Comparison on Efficiency
		3.4 Experimental Comparison on Performance
	4 Conclusion
	References
Speech Biometrics
Jointing Multi-task Learning and Gradient Reversal Layer for Far-Field Speaker Verification
	1 Introduction
	2 The Proposed Method
		2.1 Multi Tasks in the Model
		2.2 Gradient Reversal Layer
		2.3 Dynamic Loss Weights Strategy
	3 Experimental Results
		3.1 Dataset and Data Preparation
		3.2 Experiments on Different Part of the Proposed Model
		3.3 Experiments on Dynamic Loss Weights Strategy
		3.4 Comparison with the Prevailing Single-Task Models
	4 Conclusion
	References
Attention Network with GMM Based Feature for ASV Spoofing Detection
	1 Introduction
	2 Gaussian Probability Feature
		2.1 Gaussian Mixture Model
		2.2 Gaussian Probability Feature
	3 Two-Path Attention Network
		3.1 Attention Network
		3.2 Two-Path Attention Network
	4 Experiments
		4.1 Setup
		4.2 Results on ASVspoof 2019 LA Scenario
		4.3 Results on ASVspoof 2019 PA Scenario
	5 Conclusions
	References
Cross-Corpus Speech Emotion Recognition Based on Sparse Subspace Transfer Learning
	1 Introduction
	2 The Proposed SSTL Method
		2.1 Objective Function
		2.2 Optimization
	3 Experiments and Results
		3.1 Experimental Environment
		3.2 Results and Analysis
	4 Conclusion
	References
Channel Enhanced Temporal-Shift Module for Efficient Lipreading
	1 Introduction
	2 Related Work
	3 Proposed Approach
		3.1 Temporal-Shift Module (TSM)
		3.2 Squeeze-Excitation Network (SE)
		3.3 Network Structure
	4 Experiment
		4.1 Dataset
		4.2 Data Preprocessing
		4.3 Compare the Effectiveness of Attention Modules Combined with TSM
		4.4 Comparison with Some State-of-the-Arts
	5 Conclusion
	References
Explore the Use of Self-supervised Pre-trained Acoustic Features on Disguised Speech Detection
	1 Introduction
	2 Related Work
	3 Kekaimalu
	4 Experiments
		4.1 Dataset
		4.2 Statistical Moments for Traditional Features
		4.3 Feature Fusion
	5 Discussion
	6 Conclusions
	References
Author Index