Gait recognition under different clothing conditions via deterministic learning

In real-world scenarios, the clothing difference constitutes one of the most common covariate factors that can affect the performance of gait recognition systems. This paper proposes a gait recognition method which is invariant to maximum number of clothing conditions. First, four kinds of timevarying silhouette features are selected to capture the spatiotemporal characteristics of gait motion. Second, frame-to-frame gait dynamics underlying different individuals ʼ gait features is effectively modeled by radial basis function ( RBF ) neural networks through deterministic learning. Gait patterns are represented as the gait dynamics underlying time-varying gait features. This kind of dynamics information has little sensitivity to the variance between gait patterns under different clothing conditions. In order to eliminate the effect of clothing differences, the training patterns under different clothing conditions further constitute a uniform training dataset, containing all kinds of gait dynamics under different clothing conditions. A rapid recognition scheme is presented on published gait databases. Extensive experiments demonstrate the efficacy of the proposed method.