Facial age grouping and estimation via ensemble learning

2016-07-16Age estimation has been attracted lots of attention last decade. This dissertation includes six chapters. In Chapter 1, we give an introduction on this dissertation, including significance of the research, contributions of the research, and organization of the dissertation. The previous work in this area is also thoroughly reviewed. ❧ In Chapter 2, we provide the research background, which includes a brief review of related work on soft biometrics, gender recognition, race classification, age group classification and age estimation. It addresses several feature extraction methods which could be useful in representing facial aging features. Also, some classification and regression algorithms used for age grouping and estimation are discussed. ❧ In Chapter 3, we present a structured fusion method for facial age group classification. To utilize the structured fusion of shape features and surface features, we introduced the region of certainty (ROC) to not only control the classification accuracy for shape feature based system but also reduce the classification needs on surface feature based system. In the first stage, we design two shape features, which can be used to classify frontal faces with high accuracies. In the second stage, a surface feature is adopted and then selected by a statistical method. The statistical selected surface features combined with a SVM classifier can offer high classification rates. With properly adjusting the ROC by a single non‐sensitive parameter, the structured fusion of two stages can provide a performance improvement. In the experiments, we use face images in the public available FG‐NET and MORPH databases and partition them into three pre‐defined age groups. It is observed that the proposed method offers a correct classification rate of 95.1% in FG‐NET and 93.7% in MORPH, which outperforms state‐of‐the‐art methods by a significant margin. ❧ In Chapter 4, we present a novel multistage learning system, called Grouping‐Estimation‐Fusion (GEF), for human age estimation via facial images. The GEF consists of three stages: 1) age grouping; 2) age estimation within each age group; and 3) decision fusion for final age estimation. In the first stage, faces are classified into different groups, where each group has a different age range. In the second stage, three methods are adopted to extract global features from the whole face and local features from facial components (e.g., eyes, nose, and mouth). Each global or local feature is individually utilized for age estimation in each group. Thus, several decisions (i.e., estimation results) are derived. In the third stage, we obtain the final estimated age by fusing the diverse decisions from the second stage. To create diverse decisions for fusion, we investigate multiple age grouping systems in the first stage, where each system has a different number of groups and different age ranges and, thus, various decisions can be made from the second stage and will be delivered to the third stage for fusion, where six fusion schemes (i.e., intra‐system fusion, inter‐system fusion, intra‐inter fusion, inter‐intra fusion, maximum diversity fusion and composite fusion) are developed and compared. The performance of the GEF system is evaluated on the FG‐NET and the MORPH-II databases, and it outperforms existing state‐of‐the‐art age estimation methods by a significant margin. That is, the mean absolute errors (MAEs) of age estimation are reduced from 4.48 to 2.73 on FG‐NET and 3.98 to 2.91 on MORPH‐II. ❧ In Chapter 5, the GEF framework is extended. Using one single machine learning model to deal with facial age estimation has several challenges. People of different genders and age ranges tend to have different aging processes. To ponder these effects, we propose a multistage (5‐stage) learning method for facial age estimation. In the first stage—gender grouping, a model is trained to classify faces into male and female groups. For the second stage—age grouping, faces in male or female group are then classified into age groups, where each group has different age range. In the third stage—age estimation within age groups, each age group has its trained model to predict ages of faces classified in that group. For the fourth stage—fusion of decisions, based on the diversity of different decisions (i.e., age estimation results from the third stage), decisions are selected for fusion. To make fusion effective, the diversity is created by generating several age grouping systems and adopting different facial features for age estimation to obtain various decisions (i.e., estimation results). In the final stage, through error analysis on results of the previous stage, outlier faces are predicted and their estimation errors can be compensated to further improve estimation results. Our multistage learning method is verified on the FG‐NET and MORPH databases, and shows better results compared with other state‐of‐the‐art age estimation methods. ❧ In Chapter 6, a summary of the research is made and possible future research directions are discussed