Face Recognition Using Tensor Analysis

Today there are many techniques working in this field. Some of them are:

2.4.1). Face Recognition using Tensor Analysis:

Human recognition processes consider a broad spectrum of stimuli obtained from many, if not all, of the senses. The human brain is a complex system that probably applies contextual knowledge to recognize individual faces. It is futile effort to even attempt to develop a computer system using existing technologies that can closely resemble the remarkable ability of facial recognition like humans. However, the main advantage that such a system would have over a human classifier is due to the limitation of the human brain to accurately remember a large database of individuals. Over the past couple of decades, face recognition

Isometric Feature Mapping also known as ISOMAP is often used to solve dimensionality reduction problems. Some of the traditional methods for dimensionality reduction are Principal Component Analysis (PCA) and Multidimensional Scaling (MDS). However, these techniques assume that the data points lie on a linear subspace of the high dimensional input space and cannot be used to capture any inherent non-linearity of the data image. The main advantage of ISOMAP over these linear techniques and other non-linear techniques is that it is capable of efficiently calculating a globally optimal solution. It is possible for two points to be extremely close in the original data as measured by their Euclidean distances but can be extremely far apart in the lower dimensional manifold when measured by the geodesic or shortest path distances. Isometric Feature Mapping, popularly known as ISOMAP is often used to solve dimensionality reduction problems

Lades et al. developed a Gabor wavelet based face recognition system using dynamic link architecture (DLA) framework which recognizes faces by extracting Gabor jets at each node of a rectangular grid over the face image [33]. Wiskott et al. subsequently expanded on DLA and developed a Gabor wavelet-based elastic bunch graph matching (EBGM) method to label and recognize facial images [34]. Liu and Wechsler have developed a Gabor feature based classification protocol using the Fisher linear discriminate model for dimension reduction [35]. Shan et al. have developed an enhanced fisher model using the AdaBoost strategy for face recognition [36]. Zhang et al proposed a face recognition method using histogram of Gabor phase pattern [37]. . Gabor transform is used in [38] is robust for the expression, the posture and the illumination. Reference [39] tells how to constructs an eyes feature template and energy function. Those eyes location methods are used on the assumption that the face images have been exactly located, or with clear background. However, the accuracy of eyes location is affected by various factors in the