Comparing deterministic and stochastic properties of fronto-normal gait using Delay Vector Variance

Human gait is a useful biometric which has been used in many fields such as human identification, motion synthesis and clinical diagnosis. The derivation of features used to describe gait depends very much on the temporal nature of the movement signal. We focus on the fronto-normal view of gait which provides more dynamic information. Newer features based on nonlinear analyses require the establishment of nonlinear behaviour. The level of determinism of a signal also indicates what kind of auxiliary analyses may be needed. The confounding of deterministic and nonlinear properties motivates us to perform an original analysis on gait data using the recently introduced method of Delay Vector Variance. This method shows promise as it will easily indicate the level of deterministic and nonlinear behaviour of a signal separately. We look into and compare various approaches of doing this on human gait derived from video signal.