A comparison of methods to detect postural transitions using a single tri-axial accelerometer

Area of Science:

• Biomechanics
• Gerontology
• Wearable Technology

Background:

• Accurate assessment of postural transitions (PTs) is vital for understanding mobility and fall risk, particularly in aging populations.
• Existing methods for evaluating PTs can be complex or require specialized equipment.
• Wearable sensors offer a promising, accessible approach for objective mobility assessment.

Purpose of the Study:

• To develop and evaluate two novel algorithms for assessing sit-to-stand (SiSt) and stand-to-sit (StSi) transitions using tri-axial accelerometers.
• To compare the performance of algorithms optimized for chest and lower back sensor placements.
• To determine the accuracy of these algorithms in classifying PTs and estimating transition duration (TD) in both younger and older adults.

Main Methods:

• Two algorithms were developed for tri-axial accelerometer data collected from the chest and lower back.
• Algorithm 1 (chest): utilized scalar product and vertical velocity estimates.
• Algorithm 2 (lower back): employed vector magnitude and discrete wavelet transform.
• Algorithm performance was evaluated for PT classification and TD estimation against video analysis in 40 younger and 40 older adults.

Main Results:

• Both algorithms achieved excellent PT classification accuracy (>86%) across all participants.
• The chest-based algorithm demonstrated superior performance in estimating transition duration (TD).
• Intraclass correlation coefficients (ICCs) for TD estimation with the chest sensor ranged from 0.678 to 0.969 compared to video analysis.

Conclusions:

• Both developed algorithms are effective for classifying postural transitions in healthy adults.
• The chest-mounted sensor and its associated algorithm provide a more accurate method for estimating transition duration.
• These findings support the use of wearable accelerometers for objective and accurate mobility assessments, especially in older adults.