Classification of gait patterns combining their spatial and temporal properties

Gait analysis and classification is very important in medical diagnostics as well as athletic performance analysis. In this paper, we present a method combining spatial and temporal properties of features to classify different gait patterns. These features are angle data coming from gait kinematics. Mixture models are used to approximate the spatial distribution of features while stationary Markov chains describe the temporal relation within the mixture models. EM algorithm is implemented to compute parameters of mixture models and Markov chain models. The experimental results show that combination between spatial properties and temporal properties gives a good way to gait analysis.