Multimodal Integration data for Human Gait Analysis

Integration of normal gait kinematics data with video silhouettes in order to enhance markerless ana

Human gait is a complex phenomenon, it may be seen as the interaction of a control system, an energy source, levers to provide movement and forces that allow lever movement. Gait clinical analysis has been used for Medical diagnosis, Rehabilitation and Post-Surgical treatments. Typically, gait analysis is carried in laboratories that include the quantication of kinematic and kinetic patterns as well as energy consumption indicators. The importance of kinematics data, obtained using markers located in diverse body parts, is diminished due to high noise the markers and in general, the measurement technique interfere in the results. Markerless analysis techniques are promising but still have some disadvantages that have to be considered and minimized, example given, the current video processing techniques lead to leg identification mistakes or occlusion mistakes. On the other hand, kinematics of normal gait have been broadly described by researchers, thus there are detailed patterns that describe limbs´ movement in a gait cycle, and are known as kinematics patterns. Usually, kinematics patterns are used to propose mechanical models of human gait, this project will use them to obtain better results in markerless techniques. Using kinematics patterns of normal gaits a set of regular poses for a gait cycle will be obtained, a metric will be proposed, thus the relation between the silhouettes taken from the video processing techniques and the set of poses will be established.