Drift removal for improving the accuracy of gait parameters using wearable sensor systems

Area of Science:

• Biomechanics
• Wearable Sensor Technology
• Human Gait Analysis

Background:

• Wearable sensors for measuring orientation angles are susceptible to accumulated signal noise, leading to drift from true values.
• This drift introduces inaccuracies in human gait measurements, limiting the clinical utility of wearable sensor data.
• Accurate gait analysis is crucial for diagnosing and managing gait-related diseases.

Purpose of the Study:

• To propose and validate a novel method for reducing drift errors in wearable sensor data for accurate human gait measurement.
• To enhance the precision of orientation angle measurements obtained from wearable sensors.
• To demonstrate the potential of the improved method in clinical gait evaluations.

Main Methods:

• Implemented a 4th order Butterworth infinite impulse response (IIR) digital filter to denoise raw gyro sensor data.
• Subtracted the mode value of static state gyro data to eliminate offset errors.
• Employed a robust double derivative and integration technique to correct residual drift.
• Established the gravitational acceleration vector from acceleration data in standing and sitting postures to minimize sensor attachment errors.

Main Results:

• The proposed method effectively removed drift effects from wearable sensor data.
• Significant improvements in joint angle accuracy were observed: an average difference of 2.1° for hip, 33.3° for knee, and 15.6° for ankle flexion/extension compared to uncorrected data.
• Kinematic and spatio-temporal gait parameters were successfully calculated, including heel-contact and toe-off timing.

Conclusions:

• The novel method significantly reduces drift errors in wearable sensor-based gait analysis.
• The enhanced accuracy in joint angle measurement shows the method's effectiveness.
• This approach holds considerable potential for the clinical evaluation of patients with gait-related disorders using wearable sensors.