Fusion of Nonlinear Measures in Fronto-normal Gait Recognition

Human gait is an emerging biometric showing promise in its use. It incorporates time implicitly which allows a wide range of temporally based analyses to be applied. Currently, most dynamic analyses of gait employ the fronto-parallel view where people walk in a plane parallel to a camera. They employ linear signal decomposition techniques to obtain features that can be used for human recognition such as frequency and phase. The gait signal is assumed to be statistically stationary. However, most biological signals are not so well specified, many studies showing that they are nonlinear and nonstationary especially in the fronto-normal (FN) view which is more commonly encountered. We provide a novel combination of two different nonlinear measures, one exploiting chaosity and another representing regularity, which can be used to identify a person using gait. This opens up new avenues for research in gait recognition, employing nonlinear analyses on temporal features in FN gait as a biometric.