Type-2 fuzzy blended improved D-S evidence theory based decision fusion for face recognition

Abstract

Data acquisition from (single/multiple) sensors can efficiently be combined by exploiting neural networks (NN) for accurate supervising of pattern recognition especially for biometric face recognition. In case where information obtained from the different sensor shows highly conflicting, the classical Dempster’s combination rule may give rise counter-intuitive result. This paper presents type-2 fuzzy blended improved Evidence D-S (T2FIE-DSDLF) combination rule for multi sensor data fusion which minimizes paradoxes of Dempster–Shafer (D-S) combination rule. Different textural facial image (using LBP and LTP) with genetically algorithm based G2DLDA methods are employed to generate feature vector corresponding to a face image. Multiple Basic Probability Assignments (BPA) corresponding to a class with respect to different feature vectors can be combined by using Vertical Slice Centroid Type reduction (VCTR) and the mass function using evidence theory is calculated to find the score for the class. The class having the maximum score is conceived as the class of input test face image. Our T2FIE-DSDLF method is evaluated with two popular classifiers — radial basis function (RBF) neural network, and support vector machines (SVM). The numerical examples presented shows the effectiveness of our method. Moreover, to validate the efficiency of our T2FIE-DSDLF method, we have compared our method with some state-of-the-art methods.