Wearable Hip Complex Assist Robot

Assistive & Interactive Robotics Lab

AI & Robotics Institute, Korea Institute of Science and Technology, Seoul, South Korea

PI: Dr. Jongwon Lee

Achievement:

(1) Sangdo Kim, Jeonguk Kang (*: Equal Contribution), Robin Inho Kee, Sunwoo Kim, Choa Kim, Youngsu Cha, Yisoo Lee, Kanggeon Kim, Jongwon Lee, “Towards Daily Life Sarcopenia Detection: Deep Learning-Based Gait Analysis Using Wearable Hip Assistive Robot”, Under review

(2) Hobin Kim, Jongbok Lee, Sunwoo Kim, Inho Kee, Sangdo Kim, Shinsuk Park, Kanggeon Kim, Jongwon Lee, “Gait Phase Estimation Method Adaptable to Changes in Gait Speed on Level Ground and Stairs”, The Journal of Korea Robotics Society, 2023

Project: Towards Daily Life Sarcopenia Detection: Deep Learning-Based Gait Analysis Using Wearable Hip Assistive Robot

Abstract:

Daily assistive wearable exoskeletons not only support walking but also monitor individual gait patterns, providing insights into functional and health status. By extracting gait parameters that reflect physical condition, these systems enable early detection of abnormalities, intervention evaluation, and personalized feedback. In this study, we developed a deep learning-based model to estimate gait parameters using data from 22 healthy individuals and 15 sarcopenia patients. Experiments at various walking speeds in both treadmill and overground environments ensured robustness. The proposed model achieved high accuracy, with RMSE values below 7 cm for stride length, 2 cm for swing width, and 2 cm for foot clearance. Integrating the estimated gait parameters into an SVM-based classifier improved sarcopenia classification accuracy from 91.71% to 94.37%, surpassing results obtained using robotic sensor data alone. These findings highlight the effectiveness of the proposed approach and its potential for real-world and clinical applications, advancing early detection, personalized gait assessment, and adaptive rehabilitation using daily assistive exoskeletons.

What I did:

(1) Developed a deep learning model for estimating foot trajectory by fusing data from hip exoskeleton and insole sensors, achieving 100% accuracy in identifying sarcopenia patients through gait parameter analysis.

(2) Led and administered motion capture system (Motion Analysis) experiments over 40 subjects, including patients and outdoor hiking experiments over 200km.

etc.