Abstract. We present a geometric and statistical approach to gaitbased human recognition. The novelty here is to consider observations of gait, considered as planar silhouettes, to be cyclostationary processes on a shape space of simple closed curves. Consequently, gait analysis reduces to quantifying differences between underlying stochastic processes using their observations. Individual shapes can be compared using geodesic lengths, but the comparison of gait cycles requires tools for extraction, interpolation, registration, and averaging of individual gait cycles before comparisons. The main steps in our approach are: (i) off-line extraction of human silhouettes from IR video data, (ii) use of piecewise-geodesic paths, connecting the observed shapes, to smoothly interpolate between them, (iii) computation of an average gait cycle within class (i.e. associated with a person) using Karcher means, (iv) registration of average cycles using linear and nonlinear time scaling, (iv) compari...