Abstract :
In order to analyze human gait patterns, highly accurate data must be collected at high frame rates. The state of the art is to deploy a carpet-like structure instrumented with pressure sensors, which allows for measuring position, orientation and pressure of each foot at each step. Since such gait “walkway carpets” are highly expensive1 and also limited in length, we propose an alternative in the form of a wheeled walker equipped with a consumer depth camera. We have designed and implemented algorithms that derive the same set of parameters from the depth data as in a gait walkway system, however without the need for the physical presence of a walkway carpet. Moreover, we are able to provide additional information, due to continuous observation of the gait cycle, i.e. not only when the user steps on the ground. In order to retrieve actual foot pressure information, we use a shoe insole sensor. Our experiments show that the system is able to collect gait relevant data with sufficient accuracy and frame rates. While the feet´s position accuracy depends primarily on the noise of the depth sensor and is typically at a precision of less than 3 mm, the orientation accuracy is around 1-2 degrees for typical foot orientations.
