Progress in Artificial Intelligence: 20th EPIA Conference on Artificial Intelligence, EPIA 2021, Virtual Event, September 7–9, 2021, Proceedings (Lecture Notes in Computer Science, 12981)

Description

This book constitutes the refereed proceedings of the 20th EPIA Conference on Artificial Intelligence, EPIA 2021, held virtually in September 2021. The 62 full papers and 6 short papers presented were carefully reviewed and selected from a total of 108 submissions. The papers are organized in the following topical sections: artificial intelligence and IoT in agriculture; artificial intelligence and law; artificial intelligence in medicine; artificial intelligence in power and energy systems; artificial intelligence in transportation systems; artificial life and evolutionary algorithms; ambient intelligence and affective environments; general AI; intelligent robotics; knowledge discovery and business intelligence; multi-agent systems: theory and applications; and text mining and applications. Preface Organization Abstracts of Invited Speakers Responsible AI: From Principles to Action Trustworthy Human-Centric AI – The European Approach Multimodal Simultaneous Machine Translation Factored Value Functions for Cooperative Multi-agent Reinforcement Learning Contents Artificial Intelligence and IoT in Agriculture Autonomous Robot Visual-Only Guidance in Agriculture Using Vanishing Point Estimation 1 Introduction 2 Related Work 3 Visual Steering on Agriculture: The Main Approach 3.1 Hardware 3.2 Vanishing Point Detection 3.3 Autonomous Guidance 4 Results 4.1 Methodology 4.2 Base Trunk Detection 4.3 Vanishing Point Estimation 4.4 Autonomous Guidance Performance 5 Conclusions References Terrace Vineyards Detection from UAV Imagery Using Machine Learning: A Preliminary Approach 1 Introduction 2 Background 2.1 UAV Sensors 2.2 Machine Learning in Agriculture 3 Materials and Methods 3.1 UAV Data Acquisition and Processing 3.2 Dataset 3.3 Machine Learning Approach 3.4 Classifier 4 Results and Discussion 5 Conclusions and Future Work References Tomato Detection Using Deep Learning for Robotics Application 1 Introduction 2 State of the Art 3 Materials and Methods 3.1 Data Acquisition and Processing 3.2 Training and Evaluating DL Models 4 Results and Discussion 5 Conclusion References Predicting Predawn Leaf Water Potential up to Seven Days Using Machine Learning 1 Introduction 2 Background Concept 3 Materials and Methods 3.1 Experimental Field 3.2 Data Visualization and Summarization 3.3 Problem Definition and Feature Engineering 4 Experiments 4.1 Fill the Gaps 4.2 Seven Days Prediction 5 Results and Discussion 5.1 Algorithms Comparison and Variable Importance 5.2 Models Validation 5.3 Error Analysis 6 Conclusion and Future Work 6.1 Future Work References Artificial Intelligence and Law Towards Ethical Judicial Analytics: Assessing Readability of Immigration and Asylum Decisions in the United Kingdom 1 Introduction 2 Assessing Readability, and Judicial Analytics 2.1 Development and Critique of Readability Formulas 2.2 Previous Work Assessing the Readability of Legal Texts 2.3 Potential Pitfalls of Judicial Analytics 2.4 Lessons from the Literature 3 Ethical Judicial Analytics 3.1 Replicating Previous Work 3.2 Dataset and Analysis 3.3 Results 3.4 Interpretation and Critical Discussion of Results 3.5 Addressing Limitations of Standard Readability Formulas Through the Use of ML Approaches 3.6 Ethical Considerations in Judicial Analytics 4 Conclusions: Developing Ethical Judicial Analytics in Service of the Stakeholders of the Legal System References A Comparison of Classification Methods Applied to Legal Text Data 1 Introduction 2 Related Work 3 Theoretical Basis 3.1 Artificial Neural Networks 3.2 Dropout 3.3 Support Vector Machine 3.4 K-Nearest Neighbors 3.5 Naive Bayes 3.6 Decision Tree 3.7 Random Forest 3.8 Adaboost 3.9 Term Frequency - Inverse Document Frequency 4 Methodology 4.1 Type of Study 4.2 Dataset 4.3 Evaluation Measures 4.4 Machine Learning Pipeline 5 Results 6 Conclusions References Artificial Intelligence in Medicine Aiding Clinical Triage with Text Classification 1 Introduction 2 Related Work 3 Materials and Methods 3.1 Available Data 3.2 Task 3.3 Dataset 3.4 Text Representation 3.5 Experiments 3.6 Experimental Setup 4 Results 4.1 Find the ``Best'' Algorithm and Representation 4.2 Fine-Tuning the Embedding Model 4.3 Considering the Most Probable Clinical Pathways 5 Discussion 6 Conclusions References A Web-based Telepsychology Platform Prototype Using Cloud Computing and Deep Learning Tools 1 Introduction 2 Description of the System 2.1 Cloud-Based Backend Software Architecture 2.2 Web Client as Frontend 2.3 Biomedical Parameters Acquisition 3 Results 4 Conclusions and Future Work References Detecting, Predicting, and Preventing Driver Drowsiness with Wrist-Wearable Devices 1 Introduction 2 Related Work 2.1 Measurement of Driver Drowsiness 2.2 Drowsiness Detection 2.3 Drowsiness Prediction 2.4 Sleep Staging 3 Methodology 4 Results 5 Conclusion References The Evolution of Artificial Intelligence in Medical Informatics: A Bibliometric Analysis 1 Introduction 2 A Brief History of AI in Healthcare 3 Related Work 4 Methodology 5 Results 6 Discussion 7 Conclusion References Artificial Intelligence in Power and Energy Systems Optimizing Energy Consumption of Household Appliances Using PSO and GWO 1 Introduction 2 Related Work 3 Proposed Methodology 3.1 Swarm Intelligence Optimization Algorithms 3.2 Mathematical Model 4 Case Study 5 Results and Discussion 6 Conclusions References Metaheuristics for Optimal Scheduling of Appliances in Energy Efficient Neighbourhoods 1 Introduction 2 Related Work 3 Problem Definition 3.1 Representation of the Solution and Objective Function 3.2 Constraints 3.3 Search Space 3.4 Algorithms for Solving the Problem 4 Experimental Setup 5 Results and Discussion 6 Conclusion References Multitask Learning for Predicting Natural Flows: A Case Study at Paraiba do Sul River 1 Introduction 2 Materials and Methods 2.1 Study Area and Data 2.2 Streamflow Estimation Model 3 Computational Experiments 4 Conclusion References PV Generation Forecasting Model for Energy Management in Buildings 1 Introduction 2 SCADA System 3 Solar Forecasting Model 4 Case Study 5 Conclusions References Automatic Evolutionary Settings of Machine Learning Methods for Buildings' Thermal Loads Prediction 1 Introduction 2 Methods 2.1 Dataset 2.2 Machine Learning Methods 2.3 Model Selection Based on Differential Evolution 3 Computational Experiments 4 Conclusion References Artificial Intelligence in Transportation Systems Minimising Fleet Times in Multi-depot Pickup and Dropoff Problems 1 Introduction 2 Related Work 3 Preliminaries for MDPDPs 3.1 Routing Plans 3.2 Fleet Objectives 4 New Datasets for MDPDPs 5 Genetic Template for MDPDPs 6 Experiments for MDPDPs 6.1 Objective Values 6.2 Sharing Rates 6.3 Fleet Busyness 6.4 Fleet Size 7 Conclusions References Solving a Bilevel Problem with Station Location and Vehicle Routing Using Variable Neighborhood Descent and Ant Colony Optimization 1 Introduction 2 Related Work 3 Bilevel Problem: Station Location and Vehicle Routing 4 Proposed Bilevel Approach 4.1 Variable Neighborhood Descent for Station Allocation 4.2 Ant Colony Optimization for Routing Planning 4.3 Local Search Procedures and Route Selection 5 Computational Experiments 5.1 Analysis of the Results 6 Concluding Remarks and Future Works References Artificial Life and Evolutionary Algorithms Genetic Programming for Feature Extraction in Motor Imagery Brain-Computer Interface 1 Introduction 2 The Clinical Brain-Computer Interface Dataset 3 Data Preprocessing 3.1 Band-Pass Filter 3.2 Wavelet Transform 4 Sigmoid Single Electrode Energy 5 Genetic Programming 6 Proposed Single Feature Genetic Programming 7 Computational Experiments 7.1 Dimension of the Problem 8 Conclusions References FERMAT: Feature Engineering with Grammatical Evolution 1 Introduction 2 Related Work 2.1 AutoML - Automated Machine Learning 2.2 Structured Grammatical Evolution 2.3 Drug Development 3 FERMAT 4 Experimental Settings 5 Results 5.1 Feature Engineering 5.2 Absolute Performance 6 Conclusions References Ambient Intelligence and Affective Environments A Reputation Score Proposal for Online Video Platforms 1 Introduction 2 Related Works 2.1 Commercial Proposals 2.2 Academic Proposals 3 The Platform 4 Implementation 4.1 Essential Factors 4.2 Mapping Functions 4.3 Defined Metrics 4.4 Generalisation Potential and Risks 5 Conclusions and Future Work References A Reinforcement Learning Approach to Improve User Achievement of Health-Related Goals 1 Introduction 2 Proposed Model 2.1 Personal Agent 2.2 Coaching Agent 3 Results and Discussion 4 Conclusions and Future Work References Urban Human Mobility Modelling and Prediction: Impact of Comfort and Well-Being Indicators 1 Introduction 2 State of the Art 2.1 Crowdsensing Infrastructures 2.2 Well-Being and Comfort 3 Experimental Case Study 3.1 Data Collection 3.2 Data Pre-processing 3.3 Building the Models 3.4 Results 4 Discussion 5 Conclusions References Comparison of Transfer Learning Behaviour in Violence Detection with Different Public Datasets 1 Introduction 2 State of Art 2.1 RGB Based 3 Methodology and Methods 3.1 Architecture Networks 3.2 Dataset 3.3 Training Settings 4 Results and Discussion 5 Conclusion and Future Work References General AI Deep Neural Network Architectures for Speech Deception Detection: A Brief Survey 1 Introduction 2 Methodology 3 Speech Deception Detection Features 4 Deep Learning Methods to Speech Deception Detection 4.1 Long Short-Term Memory 4.2 Hybrid Networks 5 Discussions 6 Conclusions and Future Works References 3DSRASG: 3D Scene Retrieval and Augmentation Using Semantic Graphs 1 Introduction 2 3DSRASG: System Design 2.1 Block Diagram 2.2 Dataset Preprocessing 2.3 Reinforcement Learning Using Gaussian Mixture Model 2.4 Speech Processing 2.5 Text Processing 2.6 Semantic Scene Graph Generation 2.7 Scene Extraction and Enhancement 3 Results 4 Conclusion References Revisiting ``Recurrent World Models Facilitate Policy Evolution'' 1 Introduction 2 Background 2.1 Variational Autoencoders 2.2 MDN-RNN 2.3 Controller 3 Comparative Analysis 3.1 Replicating HS 3.2 Perceptual Model 3.3 Ablation Study 3.4 Training Policy 4 Conclusion and Future Work A Full Comparative Results A.1 Ablation Study Additional Results A.2 Improved Sample Policy B Model details B.1 VAE B.2 MDN-RNN B.3 Controller B.4 Hyperparameters References Deep Neural Networks for Approximating Stream Reasoning with C-SPARQL 1 Introduction 2 Background 2.1 C-SPARQL 2.2 Neural Networks for Time Series Classification 3 Methodology 3.1 Dataset 3.2 C-SPARQL Queries 3.3 Training RNNs and CNNs 4 Experiments and Results 4.1 Queries with Temporal Events 4.2 Queries with Background Knowledge 4.3 Combining Temporal Events and Background Knowledge 5 Conclusions References The DeepONets for Finance: An Approach to Calibrate the Heston Model 1 Introduction 2 Problem Formulation 2.1 Related Work 2.2 The Heston Model 2.3 The Deep Operator Networks - DeepONets 3 Method 4 Results and Discussion 5 Conclusion References Faster Than LASER - Towards Stream Reasoning with Deep Neural Networks 1 Introduction 2 Background 2.1 Laser 2.2 Neural Networks 3 Methods 3.1 Dataset 3.2 LASER Queries 3.3 Training and Testing CNNs and RNNs 4 Description of Experiments and Results 4.1 Test Case 1 4.2 Test Case 2 4.3 Test Case 3 4.4 Test Case 4 4.5 Test Case 5 5 Conclusions References Using Regression Error Analysis and Feature Selection to Automatic Cluster Labeling 1 Introduction 2 Related Works 3 Cluster Labeling Model 3.1 Step I—Definition of Attribute–Range Pairs 3.2 Step II 4 Experimental Methodology 5 Results 6 Conclusion References A Chatbot for Recipe Recommendation and Preference Modeling 1 Introduction 2 Background 3 Methodology 3.1 Intent Classification and Entity Recognition 3.2 Preference Modeling 3.3 Food Matching 3.4 Dialogue Management 4 User Validation 4.1 Results 5 Conclusions References Intelligent Robotics Exploiting Symmetry in Human Robot-Assisted Dressing Using Reinforcement Learning 1 Introduction 2 Background 3 Problem Formulation with Symmetry Based-Approach 3.1 MDP Model 3.2 Reinforcement Learning 3.3 Extending MDPs with Symmetry 4 Experimental Procedure 4.1 Experimental Setup 4.2 Kinesthetic Learning 4.3 Probabilistic Human Displacement Model 4.4 Cost Estimation 5 Evaluation and Results 5.1 Policy Learning in Simulation 5.2 Real-World Evaluation with a Robotic Platform 6 Conclusions References I2SL: Learn How to Swarm Autonomous Quadrotors Using Iterative Imitation Supervised Learning 1 Introduction 2 Methodology 2.1 Flocking Algorithm 2.2 Iterative Imitation Supervised Learning 3 A Proof of Principle of I2SL: Application to Quadrotors Control 3.1 Position of the Problem 3.2 Data Acquisition 3.3 Forward Model 3.4 I2SL Controller 4 Experimental Setting 4.1 Goals of Experiments 4.2 Baseline 4.3 Simulation Platform 4.4 Learning of the Flocking Model 5 Empirical Validation 5.1 Zigzag Experiment 6 Discussion and Future Work References Neural Network Classifier and Robotic Manipulation for an Autonomous Industrial Cork Feeder 1 Introduction 2 Implementation 2.1 System Overview 2.2 Conveyor Belt and Inspection Tunnel 2.3 Computer Vision 2.4 Cork Detection 2.5 Neural Network 2.6 Robot Arm 2.7 Gripper 2.8 Robot Arm Movement Controller 3 Results and Analysis 4 Conclusion and Future Work References NOPL - Notification Oriented Programming Language - A New Language, and Its Application to Program a Robotic Soccer Team 1 Introduction 2 Notification Oriented Programming 3 The Notification Oriented Programming Language 4 Case Study - Control of 6 Robots for the Small Size League (SSL) Category - In Simulation 5 Experimental Results 6 Conclusions References Compound Movement Recognition Using Dynamic Movement Primitives 1 Introduction 2 Recognition and Prediction Using Critical Points 2.1 Motion Recognition 2.2 Motion Prediction 3 Recognition and Prediction of Compound Movements 3.1 Motion Recognition 3.2 Motion Prediction 4 Experiments and Results 4.1 Recognition 4.2 Prediction 4.3 Accuracy 4.4 Improving the Knowledge of the Robot 4.5 Experiment with Robot 5 Conclusion References Metaheuristics for the Robot Part Sequencing and Allocation Problem with Collision Avoidance 1 Introduction 2 Related Literature 3 The Part Sequencing and Allocation Problem 4 Solving the Routing Problem 5 The Parts Sequencing and Allocation Metaheuristic Algorithm 6 Sensitivity Analysis 7 Conclusion References Knowledge Discovery and Business Intelligence Generalised Partial Association in Causal Rules Discovery 1 Introduction 2 Background 2.1 Association Rule Mining 2.2 Cochran-Mantel-Haenszel Test 2.3 Uncertainty Coefficient 3 Causal Association Rules with Partial Association and Uncertainty Coefficient 3.1 An Illustrative Example 4 Results and Discussion 4.1 Pattern Metrics Evaluation 4.2 Prediction 5 Conclusion References Dynamic Topic Modeling Using Social Network Analytics 1 Introduction 2 Related Work 3 Case Study 4 Methodology 4.1 Problem Description 4.2 Hashtag Co-occurrence Network 4.3 Stream Sampling 4.4 Community Detection 5 Experimental Evaluation 5.1 Results Discussion 6 Conclusion and Future Work References Imbalanced Learning in Assessing the Risk of Corruption in Public Administration 1 Introduction 2 Data Enrichment and Data Cleansing 3 Imbalanced Learning 3.1 Synthetic Minority Oversampling Technique (SMOTE) 3.2 Logistic Regression 4 Computational Results 4.1 Data-Level Solutions 4.2 Algorithm-Level Solutions 5 Discussion 6 Conclusions References Modelling Voting Behaviour During a General Election Campaign Using Dynamic Bayesian Networks 1 Introduction 2 Background 3 Data 3.1 Data Collection 3.2 Participants 3.3 Variables 3.4 Data Modelling 3.5 Comparison of Models 4 Results and Discussion 5 Conclusion References ESTHER: A Recommendation System for Higher Education Programs 1 Introduction 2 Literature Review 2.1 Recommendation Systems in Education 3 Recommendation System for Higher Education Programs 3.1 Requirements Specification 3.2 Knowledge Domain 3.3 Architecture 4 Preliminary Results 5 Conclusions References A Well Lubricated Machine: A Data Driven Model for Lubricant Oil Conditions 1 Introduction 2 Related Work 3 Data Collection 4 Experimental Study 4.1 Dataset 4.2 Prediction Model 4.3 Evaluation Metrics 5 Results 6 Conclusion References A Comparison of Machine Learning Methods for Extremely Unbalanced Industrial Quality Data 1 Introduction 2 Related Work 3 Materials and Methods 3.1 Data 3.2 Balancing Methods 3.3 Machine Learning Algorithms 3.4 Evaluation 4 Results 5 Conclusions References Towards Top-Up Prediction on Telco Operators 1 Introduction 2 Methodological Approach 2.1 Data Set Analysis 2.2 Building and Selecting Features 2.3 Sliding Window Regression 3 Experiments and Results 3.1 Data Set 3.2 Parameterization 4 Conclusion and Future Work References Biomedical Knowledge Graph Embeddings for Personalized Medicine 1 Introduction 1.1 Personalized Medicine 2 Methods 2.1 Knowledge Graph 2.2 Knowledge Graph Embeddings 2.3 Clustering 3 Results and Discussion 3.1 Knowledge Graph Embedding 3.2 Use Case: Prediction of Gene-Disease Associations 3.3 Use Case: Autism Spectrum Disorder (ASD) Disease Clusters 4 Conclusions References Deploying a Speech Therapy Game Using a Deep Neural Network Sibilant Consonants Classifier 1 Introduction 2 Serious Game for Sigmatism and EP Sibilant Consonants 3 Sibilants Classifier 4 Results for the Deployed Architecture 4.1 Silence Detection 4.2 PythonAnywhere Performance 4.3 CNN Optimization 5 Discussion 6 Conclusions and Future Work References Data Streams for Unsupervised Analysis of Company Data 1 Introduction 2 Non-supervised Data Analysis, Advanced Data Exploration and Visualization Tools 2.1 UbiSOM Concepts and Stream Learning Metrics 3 Experimental Setup 3.1 SOM Training 3.2 SOM Analysis 4 Related Work 5 Conclusions References Multi-agent Systems: Theory and Applications One Arm to Rule Them All: Online Learning with Multi-armed Bandits for Low-Resource Conversational Agents 1 Introduction 2 From Prediction with Expert Advice to Multi-armed Bandits 3 Related Work 4 Proof-of-Concept: Retrieval-Based Conversational Agent with Multi-armed Bandits 4.1 Finding the Best Answer Selection Criteria 4.2 Obtaining User Feedback 5 Experimental Results 6 Conclusions and Future Work References Helping People on the Fly: Ad Hoc Teamwork for Human-Robot Teams 1 Introduction 2 Notation and Background 3 Bayesian Online Prediction for Ad Hoc Teamwork 3.1 Assumptions 3.2 Preliminaries 3.3 Bayesian Online Prediction for Ad Hoc Teamwork (BOPA) 4 Evaluation 4.1 Evaluation Procedure 4.2 Metrics 5 Results 5.1 PB Scenario 5.2 ER Scenario 6 Conclusion and Future Work References Ad Hoc Teamwork in the Presence of Non-stationary Teammates 1 Introduction 2 Related Work 3 PLASTIC Policy with Adversarial Selection 3.1 Architecture 4 Experimental Evaluation 4.1 Experimental Setup 4.2 Results 5 Conclusions and Future Work References Carbon Market Multi-agent Simulation Model 1 Introduction 1.1 Carbon Tax 1.2 Multi-agent Based Simulation 2 Model Formalization 2.1 The Model 2.2 The Agents 2.3 Simulation Stages 3 Experiments and Results 3.1 Scenario 1 - Auction Market 3.2 Scenario 2 - Carbon Tax 4 Conclusions and Future Work References Cloud Based Decision Making for Multi-agent Production Systems 1 Introduction 2 Background 3 Decision Making Framework 3.1 Multi-agent System (MAS) Component 3.2 Cloud Computing Component 4 Experimentation and Deployment 4.1 Cloud-Based Decision Making 4.2 Multi-agent Based Simulation 5 Conclusion and Future Work References A Data-Driven Simulator for Assessing Decision-Making in Soccer 1 Introduction 2 Related Work 3 Simulator 3.1 Data 3.2 Models 4 Results 4.1 Simulating Sequences of Play 4.2 Building Playing Criterion 4.3 Reinforcement Learning 5 Conclusion 5.1 Future Work References Text Mining and Applications CyberPolice: Classification of Cyber Sexual Harassment 1 Introduction 2 Related Work 3 Dataset 3.1 Data Scraping 3.2 Data Cleaning and Data Labelling 4 Model Architectures 4.1 ML Classifiers 4.2 CNN Model 4.3 LSTM Model 4.4 BiLSTM Model 4.5 CNN-BiLSTM Model 4.6 ULMFiT Model 4.7 BERT Model 5 Experiment Setup 6 Results 7 Conclusion References Neural Text Categorization with Transformers for Learning Portuguese as a Second Language 1 Introduction 2 Related Work 3 Corpus 4 Transformer Models 4.1 GPT-2 4.2 RoBERTa 5 Implementation 6 Evaluation and Discussion 7 Conclusion References More Data Is Better Only to Some Level, After Which It Is Harmful: Profiling Neural Machine Translation Self-learning with Back-Translation 1 Introduction 2 Related Work 3 Methods for Back-Translation 3.1 Beam Search 3.2 Beam Search+Noise 4 Experimental Setup 4.1 NMT Architecture 4.2 Corpora 4.3 Experiments 5 Results 5.1 English German 5.2 Portuguese Chinese 6 Discussion 7 Conclusion References Answering Fill-in-the-Blank Questions in Portuguese with Transformer Language Models 1 Introduction 2 Background and Related Work 3 Experimentation Setup 3.1 Data 3.2 Models 4 Answering Fill-in-the-Blank Questions 4.1 Approach 4.2 Results 4.3 Examples 5 Answering Multiple Choice Fill-in-the-Blank Questions 5.1 Approaches 5.2 Results 5.3 Examples 6 Conclusion References Cross-Lingual Annotation Projection for Argument Mining in Portuguese 1 Introduction 2 Related Work 3 Corpora 4 Annotation Projection 4.1 Input Data 4.2 Translation 4.3 Alignment 4.4 Projection Algorithm 4.5 Intrinsic Evaluation 5 Extrinsic Evaluation 5.1 Experimental Setup 5.2 Results 6 Conclusions References Acceptance Decision Prediction in Peer-Review Through Sentiment Analysis 1 Introduction 2 Related Work 3 Materials and Methods 3.1 Evaluation Dataset 3.2 Data Preprocessing 3.3 Machine Learning Algorithms 3.4 Sentiment Analysis 4 Results 4.1 Paper Acceptance Classification 4.2 Overall Evaluation Score Prediction 4.3 Sentiment Analysis 5 Discussion 6 Conclusions and Future Work References Application of Data Augmentation Techniques for Hate Speech Detection with Deep Learning 1 Introduction 2 Related Work 3 Methodology 3.1 Datasets 3.2 Data Pre-processing 3.3 Data Augmentation 3.4 Sentence Encoding 3.5 Architectures 3.6 Experimental Setup 3.7 Measures 4 Results 4.1 Results with CNN 4.2 Results with Parallel CNN 4.3 Results with LSTM 5 Conclusions References Automated Fake News Detection Using Computational Forensic Linguistics 1 Introduction 2 Related Work 3 Resources 3.1 Corpora 3.2 Natural Language Processing Resources 4 System Description 4.1 Feature Extraction 4.2 Dataset Description 4.3 Classification Process 5 Experimental Results 5.1 Feature Analysis 6 Conclusions References Author Index