Gait recognition has received increasing attention in biometrics. However, more effort is needed to enhance the performance. In this paper, we investigate a novel descriptor for gait recognition known as Kernel-based Fuzzy Local Binary Pattern (KFLBP). The spatio-temporal static and dynamic characteristics of a gait sequence is first summarized using a Gait-Energy Image (GEI). Then, the proposed approach combines multiple FLBP with different radii to better handle uncertainty in GEI and improve the recognition performance. We evaluate the proposed method on CASIA B dataset at different view angles. We also compare the performance with other feature extraction methods and explore the impact of different walking covariates on the performance. Keywords–Gait recognition; human motion; image processing; multi-kernel; fuzzy logic; local binary patterns; gait energy image.