Schedule

Fady Alnajjar & Munkhjargal Gochoo

Title:  Personalized Robot Interventions for Autistic Children

Abstract: In this study, we are exploring robot-mediated therapy and assessment system designed for children with ASD. Utilizing the NAO humanoid robot, our system personalizes interaction sessions based on the real-time attention scores of each child. The robot's dynamic scenario customization significantly improves the quality and duration of social interactions. Our findings demonstrate that this approach not only enhances engagement and attention levels but also provides a reliable, objective method for tracking patient progress over time.

Title: Vision Transformer-Based Sign Recognition Model with Head and Hands Tunnelling Preprocessing Method

Abstract: Isolated sign language recognition is a scene and subject-invariant fine-grained video classification task, which conveys information mainly with hand gestures and facial expressions. However, these attributes are not well represented in general-purpose pre-trained Video Transformers (ViTs) due to a) deterioration of features in the interested regions when the raw frames are downsized before they are fed to the model, and b) general-purpose ViTs are not trained for the sign language domain. In our study, we presents a novel approach, namely SignNet, to enhance the sign language visual representation of transformer models by highlighting the critical regions of the head and hands with a novel preprocessing technique, termed H&HT, and exploiting transfer learning with supervised

pre-training on a large sign language dataset. The SignNet achieved the state-of-the-art top-1 accuracy of 62.82% (3-crop) and 61.81% (1-crop) on the WLASL2000 benchmark. Further, the novel model continuously excels in the revised versions of WLASL2000 and ASL-Citizen benchmarks. 

Lightning Talks

09:55, Xiao Chen, Ye Jin, Hamid Sadeghian, Abdeldjallil Naceri and Sami Haddadin - Assist-as-Needed Tele-Rehabilitation System with Web-Based Interface for Real-Time Monitoring and Adjustment (slides)

10:00, Sara Ahmadi Majd, Mohamad Rasoul Parsaeian, Mohsen Madani, Hadi Moradi and Abolfazl Mohammadi -  A machine learning web application for screening social anxiety disorder based on participants' emotion regulation (ML-SAD) (slides)

10:05, Keya Shah, Varesh Patel and Kinjal Joshi -  OpenPose, PoseNet and MoveNet: The Evolution of Deep Learning Methods in Yoga Pose Classification

10:10, Yikun Wang, Adithya Raghu, Monica Monica, Chaojun Guo, Marcela Múnera and Carlos Cifuentes - A SAR-Based Approach to Dementia Intervention Through Physical and Cognitive Interaction (slides)

10:15, J Anne Yow, Neha Priyadarshini Garg and Wei Tech Ang -  Adaptive Scooping for Assistive Feeding: Meeting User Preferences in Bite Size (slides)

Zen Koh

 Title: Rehabilitation 4.0: Transforming Neurorehabilitation through AI, Robotics, and Interoperability

Abstract: Rehabilitation 4.0 marks a transformative era in neurorehabilitation, integrating advanced technologies such as artificial intelligence (AI), robotics, and data analytics to deliver highly personalised, adaptive care. This presentation will focus on the foundational principles of Rehabilitation 4.0—consolidation, standardisation, and interoperability—highlighting how AI-driven systems and intelligent robotics are reshaping traditional rehabilitation methods. Through the use of real-time data, predictive algorithms, and automation, these innovations drive precision care, tailoring treatments to meet individual patient needs with unprecedented accuracy. We will explore the integration of human-robot interaction, smart devices, and telerehabilitation to improve patient outcomes, enhance autonomy, and boost engagement. Furthermore, the presentation will address key challenges, including ethical considerations, data privacy, and the evolving role of clinicians within this new paradigm. Rehabilitation 4.0 promises to close gaps in healthcare, offering more efficient, scalable, and effective neurorehabilitation solutions that will shape the future of patient care.

Shingo Shimoda

Title: Understanding Human Beings through the Lens of Robotics for Healthcare

Abstract:
The integration of robotics into healthcare has traditionally focused on providing assistance, enhancing efficiency, and supporting medical professionals in patient care. However, this study takes a novel approach by exploring how robotics can contribute to a deeper understanding of human beings, specifically within healthcare settings. By leveraging advanced sensing technologies, machine learning, and human-robot interaction frameworks, robotics offers unique insights into human behavior, emotions, and physiological responses that were previously difficult to capture.

This research examines how robotics, equipped with capabilities such as real-time monitoring, data analysis, and adaptive interactions, can observe and interpret human actions and conditions in healthcare environments. Through case studies and experiments involving patients with motor paralysis, we demonstrate that foundational robotics technologies play a pivotal role in enhancing our understanding of the complexities of human nature.