Lower-limb sagittal joint angles during gait can be predicted based on foot acceleration and angular velocity

Area of Science:

• Biomechanics
• Wearable Technology
• Machine Learning

Background:

• Continuous monitoring of lower-limb movement is crucial for early detection and management of diseases like orthopedic conditions.
• Calculating sagittal joint angles during daily walking is essential for risk evaluation.
• Traditional motion capture systems are impractical for everyday use.

Purpose of the Study:

• To estimate lower-limb sagittal joint angles during gait using variables measurable by wearable sensors.
• To validate the accuracy of machine learning-based estimations against motion capture data.

Main Methods:

• A feedforward neural network was developed using data from 200 healthy adults walking at a comfortable pace.
• The model utilized foot acceleration and angular velocity to estimate lower-limb sagittal joint angles.
• Data included 10 walking trials per participant to capture gait variability.

Main Results:

• The machine learning model achieved average root mean squared errors between 2.5° and 7.0° for lower-limb sagittal joint angles.
• Specific errors were: hip (7.0°), knee (4.0°), and ankle (2.5°).
• These results demonstrate the feasibility of the estimation method.

Conclusions:

• Lower-limb sagittal joint angles can be accurately estimated during gait using only foot acceleration and angular velocity norms.
• This method facilitates the calculation of joint angles during daily walking activities.
• The findings support the use of wearable sensors for continuous, real-world gait analysis and health monitoring.