Online Learning of Gait Models From Older Adult Data

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Gait Analysis

Background:

• Accurate gait analysis is crucial for assessing mobility and health in diverse populations.
• Existing methods often struggle with inter- and intra-personal gait variability, especially in older adults.
• Wearable accelerometers offer a promising avenue for unobtrusive, continuous gait monitoring.

Purpose of the Study:

• To develop and validate a novel online, individualized gait analysis approach.
• To enable adaptive modeling of gait signals for personalized analysis.
• To accurately identify key gait events in older adults using wearable sensor data.

Main Methods:

• An adaptive periodic model was developed for continuous, real-time gait signal representation.
• The method learns an individualized gait model, accommodating variations in gait patterns.
• Wearable ankle accelerometers were used to collect 6-minute walk (6MW) data from retirement home residents.

Main Results:

• The proposed algorithm demonstrated convergence within approximately four gait cycles.
• It achieved a low error rate of 3% in detecting initial swing gait events.
• The method was validated on a large dataset of older adults with various medical conditions.

Conclusions:

• The novel adaptive periodic model provides an effective approach for online, individualized gait analysis.
• This method accurately captures gait events, offering a valuable tool for assessing mobility in older populations.
• The findings support the use of wearable accelerometers for robust gait monitoring in clinical and research settings.