Note: Intro -- Preface -- Organization -- Contents -- Human Activity Recognition Using Wearable Sensors: Review, Challenges, Evaluation Benchmark -- 1 Introduction -- 2 Data-Sets -- 2.1 Data-Set Preparation -- 3 Literature Review -- 3.1 Hand Crafted Methods -- 3.2 CNN Based Methods -- 3.3 LSTM-CNN Methods -- 3.4 CNN-LSTM Methods -- 4 Proposed Hybrid Approach -- 4.1 Feature Extraction -- 4.2 NN Architecture -- 5 Experimental Results -- 5.1 Evaluation Metric -- 5.2 Experimental Set-Up -- 5.3 Results -- 6 Conclusion -- References -- Wheelchair Behavior Recognition for Visualizing Sidewalk Accessibility by Deep Neural Networks -- Abstract -- 1 Introduction -- 2 Sidewalk Accessibility Visualization -- 2.1 Proposed System -- 2.2 Related Work -- 3 Estimate Sidewalk Accessibilities -- 3.1 Dataset -- 3.2 Classifying Road Surface Conditions -- 3.3 Weakly Supervised Knowledge Extraction -- 3.3.1 Methodology -- 3.3.2 Analysis -- 3.4 Self-supervised Knowledge Extraction -- 3.4.1 Methodology -- 3.4.2 Analysis -- 4 Conclusion -- Acknowledgments -- References -- Toward Data Augmentation and Interpretation in Sensor-Based Fine-Grained Hand Activity Recognition -- 1 Introduction -- 2 Related Works -- 2.1 Human Activity Recognition -- 2.2 Generative Adversarial Networks -- 3 GAN-Based Data Augmentation -- 3.1 Data Transformation -- 3.2 Architecture Description -- 4 Experiments and Analysis -- 4.1 Implementation Details -- 4.2 Improving Classifier's Performance -- 4.3 Data Visualization -- 5 Discussion and Conclusion -- References -- Personalization Models for Human Activity Recognition with Distribution Matching-Based Metrics -- 1 Introduction -- 2 Related Work -- 2.1 User-Adaptive Models -- 2.2 Distribution Distance Metrics -- 3 Method -- 4 Experiment -- 5 Results -- 5.1 Nearest-FID-Neighbor -- 5.2 FID-Graph Clustering -- 6 Conclusion -- References. , Resource-Constrained Federated Learning with Heterogeneous Labels and Models for Human Activity Recognition -- 1 Introduction -- 2 Related Work -- 3 Our Approach -- 3.1 Problem Formulation -- 3.2 Proposed Framework -- 4 Experiments and Results -- 4.1 Discussion on Results -- 4.2 On-Device Performance -- 5 Conclusion -- References -- ARID: A New Dataset for Recognizing Action in the Dark -- 1 Introduction -- 2 Related Works -- 3 Action Recognition in the Dark Dataset -- 4 Experiments and Discussions -- 4.1 Experimental Settings -- 4.2 Frame Enhancement Methods -- 4.3 Statistical and Visual Analysis of ARID -- 4.4 Classification Results on ARID -- 4.5 Feature Visualization with ARID -- 4.6 Discussion -- 5 Conclusion -- References -- Single Run Action Detector over Video Stream - A Privacy Preserving Approach -- 1 Introduction -- 2 Related Works -- 2.1 Activity Recognition Using Wearable Sensors -- 2.2 Action Recognition in Video Data -- 2.3 Spatio-Temporal Human Action Detection -- 3 Single Run-Action Detector -- 3.1 Training Loss -- 4 Results and Evaluations -- 4.1 Results on UCF-Sports Dataset -- 4.2 Results on UR Fall Dataset -- 4.3 Real-Time Execution -- 5 Conclusion -- References -- Efficacy of Model Fine-Tuning for Personalized Dynamic Gesture Recognition -- 1 Introduction -- 2 Related Works -- 3 Method -- 3.1 User Dataset -- 3.2 Global Model Training -- 3.3 Data Augmentation -- 3.4 Personalization Strategy -- 3.5 Metric -- 4 Experiments -- 4.1 Fine-Tuned Layers -- 4.2 Training Samples -- 4.3 Batch Size and Learning Rate -- 4.4 Per-User Performance -- 5 Conclusion -- References -- Fully Convolutional Network Bootstrapped by Word Encoding and Embedding for Activity Recognition in Smart Homes -- 1 Introduction -- 2 Related Works -- 2.1 Traditional HAR Approaches -- 2.2 Deep Learning Approaches -- 2.3 NLP and TSC Coupling -- 3 Methodology. , 3.1 Problem Definition -- 3.2 NLP Encoding -- 3.3 FCN Structure -- 3.4 Sliding Window -- 4 Experimental Setup -- 4.1 Datasets Description -- 4.2 SEW Parameters -- 4.3 Networks Parameters -- 4.4 Hardware and Software Setup -- 4.5 Evaluation Method -- 5 Experimental Results -- 5.1 FCNs and LSTMs Performances -- 5.2 Training Time -- 5.3 Encoding Impact -- 6 Conclusion -- 7 Discussion and Future Directions -- References -- Towards User Friendly Medication Mapping Using Entity-Boosted Two-Tower Neural Network -- 1 Introduction -- 2 Task Definition -- 3 Method -- 3.1 Entity Boosted Two-Tower Neural Network -- 4 Experiments -- 4.1 Dataset and Evaluation Metrics -- 4.2 Experimental Details -- 5 Results and Discussion -- 5.1 Medication Clustering Result -- 6 Related Work -- 7 Conclusion and Future Work -- References -- Author Index.