Integration of Gait and Side Face for Human Recognition in Video

Combined with two-direction image matrix based principal component analysis (2DIMPCA) and multiple discriminant analysis (MDA), a new approach for human recognition is presented based on integrating information from gait and side face at the feature level. Feature extraction and dimension reducing is done to gait energy image (GEI) and side face image (SFI) respectively by 2DIMPCA, and two original feature vectors are obtained correspondingly, which are integrated into synthetic feature vectors. Then MDA is employed on the synthetic feature vectors of gait and side face to obtain fusion features vectors. Finally, the recognition process is implemented on the fusion feature vectors by nearest neighbor (NN) algorithm. The experimental results on dataset B of CASIA gait databases show that: 1) the recognition rate of the integration of gait and side face is higher than that of the single feature of gait or side face; 2) the recognition rate of the proposed approach is slightly higher and the recognition time (not including the preprocessing time) is much shorter than that of the method.