Towards unobtrusive in vivo monitoring of patients prone to falling.

Falling is a serious health problem for many elderly. To investigate whether the higher fall incidence in elderly is due to a higher probability of experiencing near falls in daily life, it is necessary to evaluate the stumble incidence of elderly in daily life. Accelerometers are already frequently used for in vivo activity monitoring. The current study investigates whether an ambulant and unobtrusive accelerometer can identify stumbles from treadmill walking using a wavelet based detection approach. Seventy nine healthy subjects walked on a treadmill with a triaxial accelerometer attached at the level of the sacrum. Stumbles were induced using a specially designed braking system (The TRiP). The TRiP evoked 30 stumbles at different phases of the swing phase. A wavelet-based detection algorithm is used to isolate the stumbles from treadmill walking, with a specificity of 99.9% and a sensitivity of 98.4%.