Statistical Measures to Test the Stability of Face Recognition Approach: Duplicating Human Faces based on Crosscorrelation Study

Abstract

This literature work suggests a novel approach for rearranging train set developed by us. The rearrangement of train set is based on cross-correlation study of facial images of individual subjects. The new approach, tried five factual human face acknowledgment techniques with standard and Senthilkumar, face databases. Further, the stability of our proposed method tested with different statistical measures such as rank, kurtosis, mean, median, mode and skewness using large face database FERET. The experimental results show that, our approach of rearranging train sets certainly improves recognition accuracy compared to traditional approaches.