Segmentation of gait cycles using foot-mounted 3D accelerometers

We describe a new gait segmentation method based on the continuous wavelet transform to identify stride-by-stride gait cycles from measurements of foot-mounted three-dimensional (3D) accelerometers. The detection of such gait cycles is indeed a crucial step for an accurate extraction of relevant gait events such as heel strike, toe strike, heel-off, and toe-off. We demonstrate the ability of this segmentation method, used in conjunction with a validated extraction algorithm, to calculate the following gait (duration) parameters for each gait cycle during the gait of a healthy young subject and of an elderly subject with Parkinson´s disease (PD) in OFF and ON states: durations of (1) loading response, (2) mid-stance, (3) push-off, (4) stance, (5) swing, (6) stride, (7) step, and (8) double support phases. The experimental results show that the proposed method can extract relevant refined gait parameters to quantify subtle gait disturbances in subjects with PD.