An Efficient Descriptor for Gait Recognition Using Spatio-Temporal Cues

In this paper, we introduce a new efficient feature for gait recognition, namely the boundary energy image (BEI). To construct this feature, the contours of the binary silhouettes in a gait sequence are extracted, and average of these contours is computed. Finally, the dimensionality of the feature set is reduced by applying PCA, and LDA is used to classify the reduced feature set. Evaluation of the proposed approach on CASIA B and TUMGAID data sets has shown significantly accurate results. Comparative study with existing techniques also show that our approach outperforms existing features like gait entropy image, active energy image and gait energy image by a substantially high margin both in terms of accuracy and response time.