Gait asymmetry analysis based on 3D volume reconstruction

Gait disorders are really complex to analyse as they require dealing with numerous parameters (i.e. all the degrees of freedom at each time step of a gait cycle). Thus, it requires complex and expensive motion capture systems which can provide the user with 3D coordinates of dozens of markers in space. These raw data have to be heavily processed in order to compute the joint angles that are necessary to calculate the indexes. This process involves computing joint centres, associating a numerical skeleton to the data, calculating the joint angles and finally returning the targeted index. All the stages of this long process involve potential inaccuracies which can lead to further complex analyses. Some of the indexes proposed in the literature deal with asymmetry and regularities in gait patterns (Hausdorff et al. 1997). Many methods have been proposed a unique index based on the numerous data provided by motion capture systems. Some authors have proposed the use of principal component analysis (PCA) to capture the main intrinsic properties of a gait pattern (Wu et al. 2007). However, it still involves joint angles. This paper aims at using few video cameras to measure gait asymmetry. The main idea is to provide small rehabilitation centres and doctors with a tool for analysing gait patterns without the need of costly motion capture systems. This method directly analyses gait along with time, using the 3D reconstructed volumes of the subject. PCA is directly applied to the reconstructed volumes to compute gait regularity and asymmetry without the need of joint angles computation. In order to validate the system, we simulate symmetrical and asymmetrical gaits along with time. These simulated gaits are rendered on a numerical mannequin filmed by virtual cameras. 2. Methods