Lower body kinematics estimation during walking using an accelerometer

Area of Science:

• Biomechanics
• Wearable Technology
• Machine Learning

Background:

• Accurate measurement of lower limb joint angles is crucial for assessing user progress in rehabilitation and sports.
• Traditional laboratory-based motion capture systems are expensive and difficult to deploy in real-world settings.
• There is a need for accessible and portable methods for quantifying joint kinematics.

Purpose of the Study:

• To continuously estimate lower limb joint angles during walking using an accelerometer and a random forest (RF) algorithm.
• To validate the accuracy of the RF-based estimation against gold-standard Vicon motion capture data.
• To explore the potential of wearable sensor technology for out-of-laboratory gait analysis.

Main Methods:

• Seventy-three subjects walked at various speeds while their lower limb joint angles were captured using a 10-camera Vicon system.
• Acceleration data from wearable sensors served as input for a random forest (RF) model to estimate ankle, knee, and hip angles.
• Statistical analysis, including Pearson correlation coefficient (r), Mean Square Error (MSE), and paired statistical parametric mapping (SPM) t-test, was used for validation.

Main Results:

• The RF model achieved high accuracy, with Pearson correlation coefficients (r) consistently above 0.91.
• Mean Square Error (MSE) values for estimated joint angles ranged from 0.04 to 24.29.
• Statistical parametric mapping (SPM) revealed no significant differences between experimental and estimated joint angles across all planes and gait cycles.

Conclusions:

• The study successfully developed an accessible and portable procedure for quantifying lower limb joint angles using accelerometers and RF.
• The wearable-based joint angle estimation demonstrates high accuracy and reliability, comparable to laboratory settings.
• This technology holds significant potential for real-world gait analysis in diverse, non-laboratory environments.