International Conference on Computer and Knowledge Engineering

Home / 11th International Conference on Computer and Knowledge Engineering

Age Estimation Based on Facial Images Using Hybrid Features and Particle Swarm Optimization

Authors :

NILOUFAR MEHRABI¹ SAYED PEDRAM HAERI BOROUJENI²

1- Istanbul Aydin university 2- Istanbul Aydin university

Keywords :

(Age Estimation, Feature Fusion, Gabor Algorithm, Local Binary Pattern Algorithm (LBP), Local Phase Quantization Algorithm (LPQ), Histogram of Oriented Gradients (HOG))

Abstract :

Face images contain many important biological characteristics. The research directions of face images mainly include face age estimation. The appearance of the face changes dynamically and these changes depend on many factors such as light, aging, makeup, beard, etc. The human face has many characteristics, including emotions, sex, race, age, etc. Taking face age estimation as an example, the estimation of face age images through algorithms can be widely used in the fields of biometrics, intelligent monitoring, commercial, military, human-computer interaction, and personalized services. For instance, one can provide content based on the age of the person in the electronics, or prevent people from reaching the age limit for purchasing cigarettes from vending machines. In general, the age estimation system is divided into four distinct stages. The first step is extracting local features. In the second step, these attributes are integrated for the Feature Fusion method in which we combined four different feature extraction methods. In the next step, the dimension of the attributes is reduced by different methods of feature selection. Finally, we used the classification and regression methods to estimate the age and age groups. We mainly used support vector machines (SVM) to classify age groups followed by support vector regression (SVR) for within age group age estimation. The errors that may happen in the classification step, caused by the hard boundaries between age classes, are compensated in the specific age estimation by a flexible overlapping of the age ranges. One of the most important issues in estimating age is the selection and extraction of features correctly. This paper uses feature extraction methods including the Gabor algorithm, Local Binary Pattern (LBP) algorithm, Local Phase Quantization (LPQ) algorithm, and Histogram of Oriented Gradients (HOG). After that, the Feature Fusion method combined the extracted features for better classification results. We also used the PSO in the proposed method to select optimal features which leads to enhance the system performance. Finally, through extensive experiments on two popular aging datasets, the FG-NET and the MORPH, we demonstrate the effectiveness of our algorithm in improving age estimation performance. We achieved an MAE of 3.34 years and 75.69% classification accuracy in the FGNET dataset, as well as an MAE of 3.21 years and 81.66% classification accuracy in the MORPH dataset.