Nonlinear optimization for drift removal in estimation of gait kinematics based on accelerometers

Area of Science:

• Biomechanics
• Sensor Technology
• Data Processing

Background:

• Accurate measurement of human movement, particularly gait, is crucial for clinical assessment and research.
• Existing methods using accelerometers face challenges with signal drift caused by time-varying and temperature-dependent offsets.
• Reliable estimation of joint angles and segment trajectories is essential for understanding biomechanical function.

Purpose of the Study:

• To develop and validate a new data processing method for estimating leg segment angles, joint angles, and sagittal plane trajectories.
• To address and correct signal drift issues inherent in accelerometer data.
• To provide a robust algorithm for real-time and off-line gait analysis.

Main Methods:

• Utilizing wireless 3D accelerometer sensor units mounted on leg segments.
• Modeling sensor signal offset using cubic spline polynomials.
• Employing nonlinear numerical simplex optimization to determine polynomial coefficients and minimize signal drift.
• Comparing estimated angles and trajectories with data from an optical motion capture system.

Main Results:

• The proposed method significantly reduces drift in processed angle and movement displacement signals.
• Root mean square errors for estimated angles were below 4 degrees.
• Errors in calculated stride length were less than 2% compared to optical motion capture.
• The algorithm demonstrated independence from gait velocity and individual gait characteristics.

Conclusions:

• The developed data processing method provides accurate and reliable estimation of gait parameters from accelerometer data.
• The technique effectively mitigates accelerometer signal drift, enhancing the usability of wearable sensors for biomechanics.
• This algorithm offers a versatile tool for both real-time and off-line gait analysis without requiring subject-specific calibration.