Multi-Resolution Local Moment Feature for GAIT Recognition

Gait recognition has recently gained significant attention from researchers, especially computer vision researchers. Compared with other biometrics, gait has its unique advantages. Other biometrics technologies, such as face recognition, hand recognition, fingerprint recognition, can´t work effectively when the person is far away. A simple and efficient gait recognition approach based on multi-resolution local moment features is proposed. For each image of gait sequence, first, it should be normalized as same center and same height. Secondly, we divide it into numbers of small blocks that have the same dimension by different methods. Thirdly, we calculate one or more features of each small block, all of them construct the feature vector of the image. Then, eigenspace transformation based on the principal component analysis (PCA) is applied to these feature vectors derived from gait sequence to reduce the dimensionality of the input feature space. Finally, SVM is used to get the correct classification rate. By utilizing the proposed approach, the experiments made on CMU database have achieved comparatively high correction identification rate