Data Knee Pads: A Lower Limb Motion Capture System Based on Heterogeneous Sensors

Area of Science:

• Biomechanics
• Sensor Technology
• Machine Learning

Background:

• Traditional motion capture systems struggle with occlusion and drift.
• Accurate lower limb motion analysis is vital for performance enhancement and user experience in fields like rehabilitation and virtual reality.

Purpose of the Study:

• To develop an advanced lower limb motion capture system using a novel data knee pad.
• To integrate Inertial Measurement Units (IMU) and liquid metal sensors for comprehensive motion and muscle activity data.
• To create a pose estimation model for accurate reconstruction of lower limb movements.

Main Methods:

• Designed a data knee pad integrating IMU and liquid metal sensors.
• Developed a pose estimation model employing Fast Fourier Transform (FFT) for signal analysis.
• Utilized Transformer encoder for feature extraction and a fully connected regression network for movement reconstruction.

Main Results:

• Achieved an average tracking error of 1.48° for lower limb pose estimation with a personalized model.
• Successfully validated the system's capability to capture muscle activity signals.
• Demonstrated high-precision measurement of knee joint bending angle alongside muscle activity detection.

Conclusions:

• The developed system offers high-precision lower limb motion capture and muscle activity monitoring.
• This integrated approach overcomes limitations of traditional systems, offering enhanced applicability.
• The technology shows significant potential for motion detection and rehabilitation evaluation, particularly for muscle-related conditions.