A Novel Approach For Improving Face Recognition Accuracy Using Fit Multi-Class Model For SVM

Facial recognition is one of the major areas within the field of artificial intelligence and image processing(pattern recognition).It has been throughly used for identity authentication,surveillance system,and biometrics.Facial recognition is a popular choice for biometrics systems because it is contact-less,natural,convenient,and generally reliable.Furthermore,the determination of accuracy in facial recognition is seen as an important area in biometric research.Consequently,the primary objective of this research is to improve upon the accuracy of facial recognition.Several methods for improving upon accuracy of facial recognition have been proposed and have been mainly based on using Support Vector Machine(SVM)techniques.Here,a novel and more efficient method of facial recognition SVM was explored.This method may significantly improve upon facial recognition accuracy and serve as a Fit Multi-Class model for SVM.In order to solve more than two class(multi-class)problems,several numbers of SVM combined multi-classification models have been employed through the combination of several binary classifiers.In the research a combined multiple-classifier is employed,namely the Classification ECOC classifier.The Histogram of Oriented Gradients(HOG)features extraction method is employed to extract related features.The Multi-class SVM is employed to classify images with the help of One-versus-One coding design method.Moreover,the ECOC is used to reduce the classification problem with multi-class to a set of binary classification problem.To prove the effectiveness of our proposed facial recognition method,the ORL,YALE face,JAFFE,and Own-created databases were used to perform experiments on facial recognition.Moreover,the obtained results were compared with those found results,which show that the proposed method is more effective than other found methods.The experimental results reveal the accuracy of our proposed method,which provided a recognition rate of over 98%.The proposed facial recognition method shows high level of accuracy.This level of accuracy has a several number of practical applications comprising crime prevention,assistive technologies for disabled,law enforcement and forensic computing.