Locomotion Mode Transition Prediction Based on Gait-Event Identification Using Wearable Sensors and Multilayer...

Area of Science:

• Biomechanics and Robotics
• Machine Learning in Healthcare
• Wearable Sensor Technology

Background:

• Daily activities involve diverse locomotion modes like walking, stair climbing, and obstacle negotiation.
• Predicting transitions between these modes is crucial for developing adaptive assistive devices, such as exoskeletons.
• Gait-event detection, including foot contact (FC) and toe-off (TO), is vital for understanding locomotion dynamics.

Purpose of the Study:

• To develop an integrated machine learning framework for simultaneous prediction of locomotion mode transitions and gait-event identification.
• To fuse data from electromyography (EMG) and inertial measurement units (IMU) for enhanced prediction accuracy.
• To enable assistive devices to anticipate and adapt to changes in locomotion modes for improved user assistance.

Main Methods:

• A two-multilayer perceptron (MLP) machine learning framework was designed.
• The first MLP was trained for gait-event detection (FC and TO).
• The second MLP was trained to predict transitions between walking, ramp ascent, and ramp descent modes using fused sensor data.

Main Results:

• The gait-event detection MLP achieved high accuracy for FC and TO, with minimal misclassifications near TO events.
• A small time difference (2.5 ms for FC, -5.3 ms for TO) was observed between predicted and true gait events.
• The locomotion mode prediction MLP demonstrated high accuracy: 96.3% for walking, 90.1% for ramp ascent, and 90.6% for ramp descent, with sufficient prediction time.

Conclusions:

• The developed framework accurately predicts locomotion mode transitions and gait events using EMG and IMU data.
• The system can anticipate transitions in the mid- to late stance phase, providing timely information for assistive devices.
• These findings hold significant potential for enhancing the performance of assistive technologies, promoting smoother mobility for individuals with motor impairments.