Performance evaluation of acceleration and jerk in unstable walking detection

Abstract

Accelerometers have been widely used for human activity recognition as an early prediction of fall risk. However, acceleration data do not consider the force of gravity. Recent studies found that jerk, the derivative of acceleration, can describe the changes of body accelerations without considering the sensor orientation. This might overcome the issues caused by the displacement of the sensor, especially if a smartphone-based accelerometer is used as the sensor. This study aimed to compare the performance of acceleration and jerk in detecting postural stability using the postural stability index (PSI). Slightly different daily activity living such as walking on a flat surface, walking upstairs, and walking downstairs were chosen to compare the sensitiveness of acceleration and jerk in detecting the slight postural sway in healthy subjects. The collected data were pre-processed using the 8-modes of ensemble empirical mode decomposition (EEMD). Then, the multiscale entropy (MSE) of each intrinsic mode function (IMF) was calculated, and in the end, the PSI values were obtained. The paired t-test calculation using acceleration data showed that walking on a flat surface and walking downstairs are significantly different (p = 0.039). Whereas, the jerk dataset could not distinguish walking on a flat surface and walking downstairs (p = 0.228). From this result, it is evident that acceleration is better in recognizing human activities than jerk.