Detection of Gait From Continuous Inertial Sensor Data Using Harmonic Frequencies

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Mobile gait analysis using wearable inertial measurement units (IMUs) is vital for assessing movement impairments in neurological and musculoskeletal diseases like Parkinson's disease (PD).
• Increasing data volumes from long-term monitoring necessitate robust and efficient analysis pipelines.
• Current gait detection methods often lack the specificity to reliably reject non-gait signals.

Purpose of the Study:

• To develop a novel algorithm for robust gait detection from continuous inertial sensor data.
• To achieve high sensitivity and specificity in distinguishing gait from other movements.
• To enable reliable long-term and free-living monitoring of gait in clinical populations.

Main Methods:

• Developed a novel algorithm for gait detection using sliding windows of IMU signals from foot-worn sensors.
• Processed signals in the frequency domain, identifying gait based on specific harmonic frequency patterns.
• Trained and evaluated the algorithm on 150 clinical measurements and validated on 203 unsupervised gait tests.

Main Results:

• Achieved a sensitivity of 0.98 and specificity of 0.96 for gait detection using angular rate around the medio-lateral axis.
• Demonstrated high reliability on an independent validation dataset with a sensitivity of 0.97 for unsupervised gait tests.
• The algorithm effectively distinguishes gait signals in both standardized and non-standardized monitoring scenarios.

Conclusions:

• The novel algorithm reliably detects gait from continuous IMU signals, addressing a key limitation in current mobile gait analysis.
• This approach shows significant promise for applications in free-living and non-standardized gait monitoring.
• Improved gait detection accuracy supports enhanced assessment and management of movement disorders.