Notions of symmetry in human movement for recognition

Abstract

Notions of symmetry are powerful for understanding as they explore relationships in nature for analysis. We describe how symmetry analysis can be used to recognize people by their gait. This approach
is reinforced by the view from psychology that human gait is a symmetrical pattern of motion and that symmetrical properties of human movement can indeed be used for human gait analysis.
Here, we use gait as a vehicle to investigate both the symmetry of moving objects as provided in our new spatial and spatio-temporal symmetry analyses. We apply these symmetry extractions to a number
of databases to demonstrate their potency. A performance analysis shows that using symmetry for gait recognition enjoys practical advantages such as relative immunity to noise, ability to handle
missing information and the capability to handle occlusion. The results show that the symmetrical properties of human gait appear to be unique and can indeed be used for analysis and for recognition with recognition rates exceeding 90%. Best performance is achieved by a spatio-temporal operator, reflecting the view that recognition by gait is not just from body shape, but also by the way the body moves.