Learning multiview face subspaces and facial pose estimation using independent component analysis

Abstract

Abstract-An independent component analysis (ICA) based approach is presented for learning view-specific subspace representations of the face object from multiview face examples. ICA, its variants, namely independent subspace analysis (ISA) and topographic independent component analysis (TICA), take into account higher order statistics needed for object view characterization. In contrast, principal component analysis (PCA), which de-correlates the second order moments, can hardly reveal good features for characterizing different views, when the training data comprises a mixture of multiview examples and the learning is done in an unsupervised way with view-unlabeled data. We demonstrate that ICA, TICA, and ISA are able to learn view-specific basis components unsupervisedly from the mixture data. We investigate results learned by ISA in an unsupervised way closely and reveal some surprising findings and thereby explain underlying reasons for the emergent formation of view subspaces. Extensive experimental results are presented.Index Terms-Appearance-based approach, face analysis, independent component analysis (ICA), independent subspace analysis (ISA), learning by examples, topographic independent component analysis (TICA), view subspaces.

Automatic head pose estimation with Synchronized sub manifold embedding and Random Regression Forests

“…The first category [12] formulates pose estimation as a conventional multiclass pattern recognition problem, and only rough pose information is inferred from these algorithms. The second category takes pose estimation as a regression problem, and nonlinear regression algorithms, e.g., Neural Network [13], are used for learning the mapping from the original appearance features to the pose label.…”

Estimating pose and illumination direction for frontal face synthesis

“…In [8,2], they are divided in five categories: shape-based geometric analysis [9], appearance-based methods [11], model-based methods [6], template-based methods [13], and dimensionality reduction based methods [2,8]. In shape-based geometric analysis, the head poses are estimated by geometric parameters defined by the facial landmarks.…”

An Empirical Study for PCA- and LDA-Based Feature Reduction for Gas Identification

“…Different research approaches have already been presented for feature reduction like independent component analysis (ICA) [12], multidimensional scaling [13], etc. The two most common techniques for dimensionality reduction are principal component analysis (PCA) and linear discriminant analysis (LDA).…”