Biased Manifold Embedding: A Framework for Person-Independent Head Pose Estimation

Abstract

The estimation of head pose angle from face images is an integral component of face recognition systems, human computer interfaces and other human-centered computing applications. To determine the head pose, face images with varying pose angles can be considered to be lying on a smooth low-dimensional manifold in high-dimensional feature space. While manifold learning techniques capture the geometrical relationship between data points in the highdimensional image feature space, the pose label information of the training data samples are neglected in the computation of these embeddings. In this paper, we propose a novel supervised approach to manifold-based non-linear dimensionality reduction for head pose estimation. The Biased Manifold Embedding (BME) framework is pivoted on the ideology of using the pose angle information of the face images to compute a biased neighborhood of each point in the feature space, before determining the low-dimensional embedding. The proposed BME approach is formulated as an extensible framework, and validated with the Isomap, Locally Linear Embedding (LLE) and Laplacian Eigenmaps techniques. A Generalized Regression Neural Network (GRNN) is used to learn the non-linear mapping, and linear multi-variate regression is finally applied on the lowdimensional space to obtain the pose angle. We tested this approach on face images of 24 individuals with pose angles varying from −90• to +90• with a granularity of 2. The results showed substantial reduction in the error of pose angle estimation, and robustness to variations in feature spaces, dimensionality of embedding and other parameters.

RPM: Random Points Matching for Pair wise Face-Similarity

“…In this section we provide results of our method on two challenging datasets, FacePix [3] and LFW [10]. We report a quantitative comparison with recent state-of-the-art methods that use a pair-wise face similarity for verification.…”

Illumination and Person-Insensitive Head Pose Estimation Using Distance Metric Learning

“…Balasubramanian et al [3] proposed the Biased Manifold Embedding (BME) framework for head pose estimation. The pose information of the given face image data is used to compute a biased neighborhood of each point in the feature space, before determining the low-dimensional embedding.…”

“…Figure 6. Other public face databases such as FERET, the CMU-PIE database,Yale Face database, and MIT database, are not used in our experiment, because none of them provide a precise measure for pose and illumination angles and also they do not contain face images with a wide variety of illumination and pose changes [3]. To assess the robustness of our approaches, we perform the experiment in two cases: pose estimation for face images without and with illumination variation.…”

“…That is, image variation due to lighting changes is more significant than variation due to different personal identities [2]. While person-independent head pose estimation has been studied reasonably well in recent years [3][4][5], robust illumination-and person-independent head pose estimation remains a challenging problem. The main contribution of this paper is to introduce a new approach to address this problem.…”

Intra-class multi-output regression based subspace analysis

“…Principal component analysis (PCA), Linear discriminant analysis (LDA) [1] aim in finding projections that are statistically significant for preserving maximum information content. Manifold learning algorithms [2] are geometrically motivated non-linear reduction methods; recently, [3] proposed a supervised manifold embedding extention. The other class of algorithms are motivated to obtain features which capture intuitive parts of objects.…”