Step Detection and Parameterization for Gait Assessment Using a Single Waist-Worn Accelerometer.

Area of Science:

• Biomedical Engineering
• Gerontology
• Rehabilitation Science

Background:

• Gait performance decline is a primary cause of independent living loss in the elderly.
• Accelerometer-based step counting is a key metric for gait assessment.
• Existing algorithms lack accuracy at low gait speeds (<0.8 m/s), limiting their use in frail older adults.

Purpose of the Study:

• To develop and validate a novel step detection algorithm for accelerometers.
• To ensure accuracy across a wide range of gait speeds, including slow and frail walking.
• To differentiate between normal and frail walking patterns for enhanced clinical insights.

Main Methods:

• A single waist-worn triaxial accelerometer was used to collect gait data from 10 healthy adults and 21 institutionalized seniors.
• A new step detection algorithm was developed and tested for sensitivity and specificity.
• Template matching and K-means clustering were employed to refine accuracy and identify distinct gait patterns.

Main Results:

• The algorithm achieved high mean sensitivity (99 ± 1%) for gait speeds between 0.2-1.5 m/s.
• Template matching reduced false positives by 73% during cycling and eliminated them in other activities.
• K-means clustering identified two distinct step patterns corresponding to normal and frail walking.

Conclusions:

• The validated algorithm accurately detects steps in older adults across various speeds, including slow and frail gaits.
• The system can aid in identifying seniors at high risk of functional decline.
• It offers a valuable tool for monitoring patient progress during exercise therapy interventions.