Identification by Gait Using Convolutional Restricted Boltzmann Machine and Voting Algorithm

Human identification by biological features has been popular in our daily life. Compared with other identification methods, human gait has specific advantages for distinguishing who is the subject. The existing gait recognition methods mainly adopt the features that designed by the hands, which is a complex feature engineering. In this paper, we propose a new gait recognition architecture based on deep learning and mathematical voting algorithm. Different with the conventional methods, the convolutional restricted Boltzmann machine (CRBM) is adopted to extract features by the unsupervised method, and the fully-connected layers are employed to learn the gait feature. Meanwhile, the voting algorithm is added to the architecture by using statistical analysis algorithm to identify subjects. The experimental results show that the proposed architecture could outperform state-of-the-arts on the CASIA-B gait dataset.