Gait analysis with multi-sensor data integration

Abstract

Feasibility assessment was performed about a gait pattern analysis based on the temporal relationships in the multidimensional data obtained from inertial measurement units (IMUs). Temporal delays between time-series data of three 6-axes accelerometer and gyroscope sensors on the left and right forefoot and over the sacrum, under normal and fast 10 m walk by 35 elderly participants. Nine temporal delay parameters were estimated in terms of maximum and significant cross-correlation in the 2-second window between all possible combinations of the time-series data, including inter-sensor parameters, suggesting applicability of the method to the whole-body analysis. One parameter successfully predicted fall history. Other parameters successfully predicted levels of need for care (AUC=0.78, accuracy=82.4% for normal walk, AUC=0.66, accuracy=76.5% for fast walk). Future studies should address whether the analysis of the detailed temporal relationships enables more accurate estimation of gait kinematics as well as gait characterization directly.