Modified Extended Kalman Filter and Long Short-Term Memory-Based Framework for Reliable Stride-Length Estimation Using Inertial Sensors.

The proposed Modified EKF framework substantially reduced stride-length estimation error compared to baseline processing.

• Absolute mean error reduced from 29.78% to 7.77%
• Standard deviation of error reduced from 20.31 to 7.17
• The improvement was achieved through wavelet-based denoising, cubic-spline interpolation, and dynamic Kalman filter gain regulation guided by acceleration zero-crossing events

LSTM models trained on Modified EKF-preprocessed data achieved superior stride-length estimation performance.

• Mean Absolute Error (MAE) of 0.0376 was achieved
• Coefficient of determination (R²) of 0.7066 was achieved
• LSTM networks were used as the downstream learning component after the Modified EKF preprocessing pipeline

Experimental data were collected from twelve healthy participants using both inertial sensors and motion capture as ground truth.

• Twelve healthy participants total: seven females (mean age 26.76 ± 3.01 years) and five males (mean age 25.81 ± 1.63 years)
• Participants walked at self-selected speeds on a treadmill
• Both an inertial sensor-based gait monitoring system and a motion capture system were used simultaneously
• The motion capture system served as the ground-truth reference for stride-length measurements

The processing pipeline incorporated dynamic Kalman filter gain regulation to mitigate transient errors around abrupt turning points.

• Acceleration zero-crossing events were used to guide dynamic gain regulation in the Kalman filtering stage
• The modification was specifically designed to address transient errors around abrupt turning points
• Front-end preprocessing included wavelet-based denoising and cubic-spline interpolation prior to the Kalman filtering stage

Gait analysis using wearable inertial sensors faces challenges due to noise contamination and signal variability that degrade spatiotemporal measurement accuracy.

• Robust gait characterization is described as 'challenging due to noise contamination and variability in sensor-based signals'
• Accurate spatiotemporal measurements are identified as essential for early intervention, particularly in aging populations and clinical screening contexts
• The study targets applications in clinical diagnostics, rehabilitation monitoring, and health management