Hip Exoskeleton:Motion Recognition Based on Deep learning

This code is significant in achieving high accuracy motion recognition using only an encoder through AI

Project Name

Hip Exoskeleton : Motion Recognition Algorithm Based on Deep Learning by Encoder

AJOU University, Mechatronics Engineering

AML

Project Abstract

The hip-based exoskeleton provides appropriate torque during tasks to effectively reduce the muscle load on wearers, as demonstrated through electromyography (EMG) sensor-based human performance evaluation. For instance, the assistive torque for lifting tasks is 5-10 times stronger than the torque for walking joints. To deliver appropriate torque for various tasks, motion recognition algorithms are employed. Traditional exoskeletons use multiple sensors and encoders for motion recognition, which increases the load and can interfere with task performance. Moreover, existing exoskeletons, which rely on real-time angle measurement, struggle with fast-changing hip movements, leading to delayed assistive torque and reduced performance. To address these issues, a deep learning-based hip motion recognition algorithm is proposed, utilizing only encoders to quickly recognize task and walking motions and provide suitable torque, thereby reducing the load and enhancing performance.