Step Length Estimation with Wearable Wrist Sensor using ANN.

Area of Science:

• Biomechanics
• Gait Analysis
• Wearable Sensor Technology

Background:

• Step length is a crucial metric for gait analysis and assessment.
• Current dynamical models for wrist-based step length estimation using recursive techniques are insufficient.
• Accurate gait assessment requires reliable methods for determining step length.

Purpose of the Study:

• To develop and validate a novel method for estimating step length using wrist-worn inertial measurement unit (IMU) data.
• To address the limitations in current dynamical models for gait analysis.
• To improve the accuracy of step length measurement in diverse populations.

Main Methods:

• Utilized angular velocity data from a wrist sensor.
• Employed an artificial neural network (ANN) to map wrist angular velocity dynamics to thigh dynamics.
• Applied an unscented Kalman filter (UKF) to determine foot-to-hip horizontal position and subsequently estimate step length.

Main Results:

• The proposed method demonstrated an average accuracy of 81.8% for step length estimation in young individuals.
• The method achieved an average accuracy of 91.1% for step length estimation in elderly individuals.
• Results were validated against a reference system using treadmill data.

Conclusions:

• The developed technique effectively estimates step length using wrist-worn IMU data.
• This method offers a promising, non-invasive approach for gait analysis.
• The high accuracy in both young and elderly populations suggests broad applicability.