Selecting biomechanical variables for detect gait events using computational vision

Abstract

In clinical gait analysis gait events, such as initial contact and toe off, are needed for biomechanical data normalization and for several temporal/distance parameters estimation. Therefore, there is a major necessity to identify the key gait events accurately. In this paper it will be presented an algorithm to detected stance phase and swing phase events using just visual information acquired by image cameras. The performance of our algorithm was compared with the golden standard method (based on data from force platforms) considering normal video digital cameras and high-speed cameras also.