Human age-estimation system based on double-level feature fusion of face and gait images

摘要

Age estimation is one of the latest research topics nowadays in the field of image processing and computer vision with variety of commercial and security applications. To the best of our knowledge, no work has been done yet to estimate the age using fusion of two biometric traits. In this article, we propose a double-level feature fusion method to artificially estimate the age of a human being. In the first level fusion, features of several walking periods of Gait sample and features of several angles of a face sample are fused individually using averaging function. In the second-level fusion, these individually fused features of gait and face are combined to form a single-feature vector using concatenation. Information fusion from various sources increases the certainty in decision-making. Gait is a good feature to utilise as it captures the movement of the whole body and encodes both structural and transitional motion. The proposed double-level feature fusion approach improves the accuracy of the estimated age in comparison to the age estimated using single biometric trait. Further, robustness and applicability of the system for the main task of age estimation also improve to a noticeable extent.