Estimation of Lower Extremity Muscle Activity in Gait Using the Wearable Inertial Measurement Units and Neural...

Area of Science:

• Biomechanics
• Wearable Technology
• Machine Learning

Background:

• Inertial measurement units (IMUs) are increasingly used in gait analysis but only capture segment kinematics.
• Assessing muscle behavior, crucial for comprehensive gait analysis, typically requires resource-intensive methods like electromyography (EMG) or musculoskeletal modeling.
• There is a need for accessible and less demanding methods to evaluate muscle activity in gait analysis.

Approach:

• This study investigates the use of neural networks (NN) with IMU data to estimate the activity of nine lower extremity muscles.
• Two NN architectures were developed and compared: a feedforward neural network (FNN) and a long short-term memory (LSTM) network.
• The research focuses on leveraging IMU sensor data for muscle activity prediction.

Key Points:

• Both FNN and LSTM networks demonstrated capability in predicting muscle activities from IMU data.
• The LSTM network exhibited superior performance compared to the conventional FNN.
• The findings validate the feasibility of using IMU data and NN for muscle activity estimation.

Conclusions:

• This research confirms that muscle activity can be effectively estimated using IMU data and neural networks.
• The proposed method offers a potential pathway for conducting gait analysis outside traditional laboratory settings.
• This approach allows for gait analysis with a minimal number of devices, enhancing accessibility and practicality.