Human Posture Detection Method Based on Wearable Devices

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Wearable Technology

Background:

• Dynamic human motion detection is crucial for applications like motion capture and rehabilitation engineering.
• Integrating multimodal sensor data offers a comprehensive approach to understanding human movement.
• Wearable technology enables real-time, non-invasive monitoring of physiological and kinematic signals.

Purpose of the Study:

• To propose effective virtual driving control and gait recognition methods using multimodal wearable sensor data.
• To develop an effective wearable system for dynamic human posture detection.
• To assess the accuracy and feasibility of the proposed methods in real-time applications.

Main Methods:

• Utilized surface electromyography (sEMG) signals from the upper limb and triaxial acceleration/plantar pressure signals from the lower limb.
• Employed moving average window and threshold comparison for sEMG segmentation, extracting features like standard deviation and wavelet coefficients.
• Applied an optimized support vector machine (SVM) algorithm for classification and identification of actions and gait patterns.

Main Results:

• Achieved an average identification accuracy of 90.90% for three virtual driving actions using sEMG data.
• Attained an average accuracy of 90.48% for gait identification by combining triaxial acceleration and plantar pressure features.
• Demonstrated the capability of dynamically detecting motion posture information through various combinations of upper and lower limb wearable sensors.

Conclusions:

• The proposed multimodal approach effectively enables dynamic human motion detection.
• The developed methods show significant potential for applications in virtual rehabilitation systems and walking assistance devices.
• Integration of wearable sensors on different body parts enhances the accuracy and scope of human motion analysis.