Keynote Speaker_ICIET 2026

Takumi Miyoshi, Japan

Biography: Takumi Miyoshi received the B.E., M.E., and D.E. degrees in electronic engineering from the University of Tokyo, Tokyo, Japan, in 1994, 1996, and 1999, respectively. He is presently a professor at College of Systems Engineering and Science, Shibaura Institute of Technology, Saitama, Japan. He was a research associate at Global Information and telecommunication Institute, Waseda University, from 1999 to 2001, and a research fellow at Telecommunications Advancement Organization of Japan from 1998 to 2003. He was a visiting scholar at Laboratoire d'Informatique de Paris 6 (LIP6), Sorbonne University, Paris, France, from 2010 to 2011. His research interests include multimedia communication, overlay networks, sensor networks, machine learning, smart city, and digital twin. He is a member of IEEE and a senior member of IEICE.

Talk title: Toward Scalable Urban Digital Twins: Integrating Point Cloud Sensing and Real-Time Traffic Simulation

Absract: Smart cities are emerging as a new paradigm enabled by the integration of cyber and physical spaces, where digital twins play a key role in supporting data-driven urban services. This talk presents an approach to constructing spatial digital twins using 3D point cloud data collected through mobile crowdsensing (MCS) with LiDAR-equipped devices. The collected data are efficiently integrated through registration techniques while addressing challenges such as noise, varying density, and data reliability. We also introduce a traffic digital twin framework using the CARLA simulator, where real-world vehicle trajectories are utilized to reproduce dynamic traffic conditions. By combining sensing-based spatial digital twins and simulation-based traffic digital twins, we demonstrate a unified approach for scalable urban digital twin construction.

National Chung Hsing University, Taiwan

Biography: Chih-Peng Fan received the B.S., M.S., and, and Ph.D. degrees, all in electrical engineering, from National Cheng Kung University, Taiwan, in 1991, 1993 and 1998, respectively. During October 1998 to January 2003, he was a design engineer at Computer and Communications Research Laboratories (CCL), Industrial Technology Research Institute (ITRI), Hsinchu, Taiwan. In 2003, he joined the Department of Electrical Engineering at National Chung Hsing University in Taiwan as an Assistant Professor. He became a full Professor in 2013. He has more than 110 publications, including technical journals, technical reports, book chapters, and conference papers. His teaching and research interests include deep-learning based digital image processing and pattern recognition, digital video coding and processing, digital baseband transceiver design, VLSI design for digital signal processing, and fast prototype of DSP systems with FPGA and embedded SOC platform. He served and is serving as: General Chair of ICCE-TW 2018; Executive Conference Chair of ICCE 2024; IEEE Transactions on Consumer Electronics -Associate Editor (2022-now); Member of Editorial Board of Journal of Real-Time Image Processing; IEEE CTSoc Representative at the IEEE Systems Council's AdCom (2020-2021); IEEE CTSoc Representative at the IEEE Sensors Council's AdCom (2023-Now); Chair of IEEE CTSoc Sensors and Actuators (SEA) TC (2025-Now); He is a member of Taiwan IC Design Society (TICD), IEICE, and IEEE.

Talk title: OpenPose Based Yoga Exercise Learning Assistant Design with User–Instructor Synchronization and Pose Difficulty Evaluation Technologies for Dynamic and Static Yoga Self-Practice Assistant System on the GPU-Based Platform

Absract: Yoga is popular across all age groups because of its benefits for physical and mental health. To assist beginners with yoga self-practice. In this keynote talk, the OpenPose based yoga self-practice assistance system for dynamic and static yoga by angle-based poses matching and pose difficulty estimation on the NVIDIA Jetson Nano platform is introduced. The developed system uses the OpenPose Body25 model to extract the important information related to body key joints by evaluating the accuracy of users’ poses against yoga instructor demonstrations with user-instructor synchronization on the basis of joint angle differences and total distance between adjacent video frames, and then the user’s feedback with fuzzy based scoring strategy will be processed simultaneously. To prevent overly difficult yoga poses from causing injuries to beginners, the developed system includes a difficulty assessment feature that allows users to select poses according to their ability. The proposed system evaluates pose difficulty from the front and side views by estimating angular velocity, body area, body bending direction, flexibility requirements, and range of motion on the basis of joint angles and vectors. Then the information of developed Yoga difficulty estimation levels is integrated into the developed yoga self-practice system. The developed system’s effectiveness was validated on over 100 yoga pose images obtained from different sources. Strong correlations were observed between the ratings provided by the system and instructor, confirming the accuracy of the system and its potential for improving safety and adaptability in yoga self-practice.