A Structure-Based Human Facial Age Estimation Framework under a Constrained Condition

A Structure-Based Human Facial Age Estimation Framework under a Constrained Condition

Abstract:

Developing an automatic age estimation method towards human faces continues to possess an important role in computer vision and pattern recognition. Many studies regarding facial age estimation mainly focus on two aspects: facial aging feature extraction and classification/regression model learning. To set our work apart from existing age estimation approaches, we consider a different aspect - system structuring, which is, under a constrained condition: given a fixed feature type and a fixed learning method, how to design a framework to improve the age estimation performance based on the constraint? We propose a four-stage fusion framework for facial age estimation. This framework starts from gender recognition, and then go to the second phase, gender-specific age grouping, and followed by the third stage, age estimation within age groups, and finally ends at the fusion stage. In the experiment, three well-known benchmark datasets, MORPH-II, FG-NET, and CLAP2016, are adopted to validate the procedure. The experimental results show that the performance can be significantly improved by using our proposed framework and this framework also outperforms several stateof- the-art age estimation methods.

Existing System:

In the previous several years, human facial age estimation has attracted a lot of interests in the computer vision society because it can be used in many important and practical applications, such as age-related information retrieval, human-computer interaction (HCI) biometrics, internet access and security control information retrieval related problems, hashing (i.e., binary code learning) based methods have been used to solve multimedia retrieval and classification tasks. Two recently proposed hashing methods, dubbed discrete proximal linearized minimization (DPLM) and discrete cross-modal hashing (DCH), are shown to achieve state-of-the-art results for retrieval problems.

 

Disadvantage:

If computers could offer correct age estimates on human faces, for example, then computers can automatically adjust the text size on screens based on the ages of users and vending machines can determine whether to allow the customers to purchase tobacco/alcohol or not. So far most age estimation approaches use a single machine learning model trained based on some features to predict ages of the faces.

Proposed System:

Our proposed framework can offer higher flexibility than other methods, which means an age estimation method can be incorporated into our framework and have its performance improved. Take the fusion stage for example, age estimation results from a method can be treated as one decision and be added to our decision set. The larger and diversified decision set we have, the better performance the framework could offer.

To the best of our knowledge, this is the first work to investigate an age estimation problem under a constrained condition, which is a totally different aspect. Under the constrained condition, we fully investigate 252 age grouping and estimation systems. We also show the age grouping accuracy of a system and its corresponding age estimation MAE (mean absolute error) on the experimental results. This will help one to determine which system to choose if only one individual system is needed.