Feature Extraction Using Uniform Pattern based Standard Deviation for Face Recognition
Md. Abdullah Al Mamun, Md. Ariful Islam, Masud Ibn Afjal, Md. Abdulla Al Mamun, U. A. Md. Ehsan Ali, Md Palash Uddin · 2018
Feature extraction technique is an important task in efficient face detection and recognition application. A good feature extraction technique incorporates the classifier to improve the classification accuracy and to reduce the computing time. The feature extraction techniques such as Principle Component Analysis (PCA) reduces the number of dimensions without much loss of information and Histogram Oriented Gradient (HOG) counts occurrences of gradient orientation in localized portions of an image. As opposed to them, this paper introduces a feature extraction technique named Uniform Pattern based Standard Deviation (UPSD) which is based on Standard Deviation on a matrix pattern. Those patterns are needed to be uniformed by adding some priority. The experimental result shows that the classification accuracy for the real face image dataset is 85.19%, 88.89% and 92.59% using HOG, PCA and the proposed UPSD feature extraction approaches respectively. For ORL face dataset, both PCA and the proposed UPSD feature extraction approaches show the identical accuracy 91.25%.