Gait Recognition With Skeleton Information By Using Ensemble Learning

In recent years, the demand of robust gait recognition for practical applications has increased and the existing gait recognition rate still has room to improve. So an innovative Kinect-based method of gait recognition in the skeleton model by using center of mass (CM) and ensemble learning is presented to get the improvement in this paper. CM-related static features are chosen automatically by filtering selection algorithms and CM-related dynamic features are processed by spectrum analysis. This behavior preserves the better features, further improves the recognition effect and enhances the difference among features. The categories and dimensions of features lead to difficulties in the measurement of the skeletal model. To tackle this problem, methods of combining Dynamic Time Warping (DTW) with some classifiers are proposed. In order to optimize the recognition effect, ensemble learning is used to integrate the learning results of each classifier. Experiments show that the performance of our method is better than other state-of-the-art methods.