Feature Disentangling Machine - A Novel Approach of Feature Selection and Disentangling in Facial Expression Analysis

Abstract

Abstract. Studies in psychology show that not all facial regions are of importance in recognizing facial expressions and different facial regions make different contributions in various facial expressions. Motivated by this, a novel framework, named Feature Disentangling Machine (FDM), is proposed to effectively select active features characterizing facial expressions. More importantly, the FDM aims to disentangle these selected features into non-overlapped groups, in particular, common features that are shared across different expressions and expression-specific features that are discriminative only for a target expression. Specifically, the FDM integrates sparse support vector machine and multi-task learning in a unified framework, where a novel loss function and a set of constraints are formulated to precisely control the sparsity and naturally disentangle active features. Extensive experiments on two well-known facial expression databases have demonstrated that the FDM outperforms the state-of-the-art methods for facial expression analysis. More importantly, the FDM achieves an impressive performance in a crossdatabase validation, which demonstrates the generalization capability of the selected features.

Doubly Sparse Relevance Vector Machine for Continuous Facial Behavior Estimation

“…Then the performance slowly decreases for larger number of patches. This confirms previous work [22], [57], which found a sparse patch subset to be sufficient for recognizing FAUs and that the inclusion of further information leads to lower performance. …”

“…[19], [23], [51]). The only exceptions include the part-based methods for detecting facial actions units (FAUs) [20], [41], [56], those for classifying basic emotion categories [22], [53], [57], and those for pain classification [21], [27]. However, these approaches are not suitable for our problem due to the following limitations.…”

“…[22] combines sparse linear SVM with multitask learning for recognizing the six basic emotions. They learn sparse emotion-specific and shared sets of feature dimensions.…”

Active AU Based Patch Weighting for Facial Expression Recognition

“…For the CK+ database, the proposed algorithm achieves the highest accuracy of 94.09%. As we are focused on seven class expression recognition, those work developed for six expressions like [21,37,41,[43][44][45]49] are not included for comparison in this paper.…”

Is 2D Unlabeled Data Adequate for Recognizing Facial Expressions?

“…These features can be either hand-designed or learned from the training data. It is known that some features are more critical for analysing facial expressions than the others and the feature selection procedure can be applied to improve the performance [3], [4]. Indeed, extracting complex 2D or 3D features can improve the systems performance, but often requires more computational resources.…”