Face Recognition Model Based on Covariance Intersection Fusion for Interactive devices

Face recognition has become more important recently than before. The objective of this study was to develop a face recognition model that can achieve high recognition accuracy. The better the extraction of features, the better results achieved. The already existing face recognition systems suffer from the problem of integration of different types of features. In this paper, a fusion feature-level face recognition method (FFLFRM) is proposed. The input face image is detected by applying the Haar-cascade method. After that, the features are extracted by using two statistical methods: local binary pattern (LBP) and principal component analysis (PCA). Then, the Covariance Intersection Fusion (CIF) technique is applied to integrate the LBP-, and PCA-extracted features. Afterwards, the integrated feature vector is input to a Multi-layer Perceptron Artificial Neural Network (MLPANN). Performance of the proposed method was tested using the Olivetti Research Laboratory (ORL) database of face images. To assess robustness of the proposed FFLFRM, it was applied on images with and without illumination and change in pose; and images with different expressions, occlusions, and levels of image resolution. For validation purposes, the proposed method was compared with a method applying LBP only, PCA only, and a combination of LBP and PCA with Frequency Partition (FP. Performance evaluation uncovered that the proposed FFLFRM has an average recognition accuracy of 96.75%. The recognition accuracy of this proposed method is quite good, particularly when compared with the corresponding accuracies reported by other studies that used the same face images