Coefficient of Variation based Decision Tree Classifier for Face Recognition with Invariant Moments
Satya Sreedevi Redla, Banitamani Mallik, Vamsi Krishna Mangalampalli · 2020
A biometric facial recognition system helps to identify individuals based on their distinctive physiological facial features. Moreover, the biometric identifiers have been more capable and consistent than knowledge or token based techniques in recognizing individuals. The main concept of our work is to develop a facial recognition approach which increases accuracy of classification and identification using a class of decision trees based on coefficient of variation gain as splitting criteria. The body of the paper discusses about the Coefficient of Variation based decision tree classifier (CVDT) using 7 Hu invariant moments, Mean, S.D and CV as feature measures. After discretization and extracting reduct feature set, the proposed criterion classification accuracy was found to be 91%. For the sake of illustration, the proposed system is defined over a random sample of 10 frontal faces chosen from bio id data base.