Predicting lower limb joint kinematics using wearable motion sensors.

Area of Science:

• Biomechanics
• Wearable Technology
• Gait Analysis

Background:

• Estimating sagittal plane kinematics of the ankle, knee, and hip during gait is crucial for understanding lower limb biomechanics.
• Traditional motion capture systems using reflective markers are accurate but lack portability and are expensive.
• Developing wearable systems for gait analysis offers a promising alternative for clinical and research applications.

Purpose of the Study:

• To estimate sagittal plane ankle, knee, and hip gait kinematics using 3D angular velocity and linear acceleration data from motion sensors.
• To evaluate the accuracy of intra-subject and inter-subject predictions.
• To assess the impact of sensor data loss on prediction accuracy.

Main Methods:

• Collected kinematic data using reflective markers for hip, knee, and ankle joints.
• Simultaneously recorded foot and shank angular velocity and linear acceleration using integrated sensors.
• Employed a generalized regression networks algorithm to predict kinematics from sensor data.

Main Results:

• Intra-subject predictions demonstrated high accuracy with correlations of 0.93-0.99 and mean absolute deviations <= 2.3 degrees.
• Inter-subject predictions yielded lower correlations (0.70-0.89) and larger angle differences (4.91-9.06 degrees).
• Angular velocity data and shank sensor data provided minimal additional accuracy.

Conclusions:

• A wearable system using only footwear-mounted sensors and acceleration data shows potential for gait analysis.
• Predictions remained generally stable despite sensor data loss.
• Further research is needed to confirm the robustness of the generalized regression networks algorithm for activities beyond level walking.