Gait recognition based on LLE

In order to reduce dimensionality and select parameters in LLE for gait recognition, two methods are discussed in the paper. One is to train samples in same class and recognize the identification; the other is to train samples in all classes, then recognize the identification. The experimental results indicate the maximal correct classification rates (CCR) for two methods are alike when the right parameters are selected. When the parameter of neighbor points is selected near the sample number in same classification and dimensionality in lower space for LLE is chose at right value, the correct classification rate is higher in both methods.