To identify optimal kinematic signals for reliable use as inputs to machine learning-based gait monitoring systems (without extensive data processing), we quantified the level of changes in two kinematic signals around three axes in four locations. Wearing inertial motion unit wearables (IMU), 30 typically developing children (8-18yrs) walked on treadmill & outdoor overground at three different speeds giving a sizable normative dataset. Primary outcome measures were curve-based similarity analysis (specifically, cosine, Euclidean distance, Poincare and a newly defined Bilateral Symmetry Dissimilarity Test, BSDT) between treadmill and outdoor over-ground walking. Similarity analysis showed a distinct previously unreported high/middle/bottom banding pattern and superior-inferior shank acceleration (SI shank Acc) and medial-lateral shank angular velocity (ML shank AV) demonstrated the least variability across the different walking conditions (as measured by the BSDT). As secondary outcomes measure, conventional spatiotemporal gait parameters (parameter-based similarity analysis) were measured and showed varying differences across walking speeds consistent with previous literature.
|