Discriminant Incoherent Component Analysis

Abstract

Abstract-Face images convey rich information which can be perceived as a superposition of low-complexity components associated with attributes, such as facial identity, expressions and activation of facial action units (AUs). For instance, low-rank components characterizing neutral facial images are associated with identity, while sparse components capturing non-rigid deformations occurring in certain face regions reveal expressions and action unit activations. In this paper, the Discriminant Incoherent Component Analysis (DICA) is proposed in order to extract lowcomplexity components corresponding to facial attributes, which are mutually incoherent among different classes (e.g., identity, expression, AU activation) from training data, even in the presence of gross sparse errors. To this end, a suitable optimization problem, involving the minimization of nuclear-and 1-norm, is solved. Having found an ensemble of class-specific incoherent components by the DICA, an unseen (test) image is expressed as a group-sparse linear combination of these components, where the non-zero coefficients reveal the class(es) of the respective facial attribute(s) that it belongs to. The performance of the DICA is experimentally assessed on both synthetic and real-world data. Emphasis is placed on face analysis tasks, namely joint face and expression recognition, face recognition under varying percentages of training data corruption, subject-independent expression recognition, and action unit detection by conducting experiments on 4 datasets. The proposed method outperforms all the methods that is compared to in all tasks and experimental settings.

The Conflict Escalation Resolution (CONFER) Database

“…Furthermore, some researchers are critical of the grid-based feature extraction, suggesting that the sub-regions are not necessarily well aligned with meaningful facial features [53]. Motivated by these findings and other recent works [54,33,55], in this study we adopt a hybrid approach to appearance feature extraction. In particular, we first apply the same transformation used for point registration to the pixel intensities of each face image to remove translation, scale and in-plane rotation effects.…”

“…Moreover, due to the involved parties being engaged in naturalistic competitive conversations, the subjects often perform abrupt and extreme head movements (e.g., head nods, shakes, tilts), body movements (e.g., forward/backward leaning, spinning periodically on their swivel chairs) and gestures (e.g., hand crosses, hand wags). The aforementioned conditions pose obstacles to the computer vision pre-processing tasks, such as face detection, facial point tracking and registration [30,31], since the latter have to cope with frequent and large out-of-plane head rotations and occlusions [32,33]. Annotations.…”

Masked Linear Regression for Learning Local Receptive Fields for Facial Expression Synthesis

“…However, such methods require multiple facial landmarks to be pre-identified and do not propose a unified framework for dealing with hidden information, such as teeth, that is usually added in a separate, post-processing step. For the case of learning-based techniques, FES has received relatively less attention compared to expression recognition or face recognition across varying expressions (Zeng et al 2009;Jain and Li 2011;Georgakis et al 2016). Cootes et al (2001) combined shape and texture information into an Active Appearance Model (AAM).…”

Side Information in Robust Principal Component Analysis: Algorithms and Applications

“…The least resulting reconstruction residual thereupon determines its identity or expression. We refer readers to [12] for the exact problem set-up and implementation details. Table 2 collects the computed recognition rates.…”