A Comparative Study of Age Estimation Using Edge Detection and Regression Algorithm

Abstract

Age Prediction has received a lot of attention over the years because of its numerous applications ranging from
anthropology to archaeology and even forensic science. Even though there have been many methodologies
developed for this purpose, it is still a problem owing to the high variability in physiological age indicators. This
study compares different combinations of Edge Detection and Regression algorithms to determine the best
possible way to predict the apparent age of an individual using the Histogram of Oriented gradients to extract
features from the Detected Edges. The FGNET Dataset which contains over 1000 images of people of ages ranging
from 0 to 70 years old with each individual represented at least 4 different ages was used. The Edge Detection
algorithms used were Canny Edge and Sobel Filter combined with the Support Vector Regression and K-Nearest
Neighbour Regression Algorithms. The performance of the Canny edge detection algorithm and the Sobel filter
when combined with the HOG feature extraction algorithm were compared. It was observed that the combination
of the Canny Edge Detection Algorithm and the Support Vector Regression Algorithm gave the best Predictive
Accuracy