Gender recognition based on fusion on face and gait information

This paper considers the combination of face and gait biometrics from the same walking sequence to carry out gender recognition. A camera is capturing the side view of a person, while another camera is placed to record the face of the same person at the front view. After these videos are acquired, we extract the silhouette images from the gait videos and normalized frame images decomposed from the face videos. Then, for face classification, we introduce PCA to reduce the image dimension and SVM to classify gender, for gait classification, we divide the silhouette into seven parts and extract features from each and also employ SVM to classify gender. On the decision level, the sum rule is applied to implement the fusion of these two classification results. The final fusion results show an improvement on correct classification rate.