Robust facial expression recognition via compressive sensing

Abstract

Recently, compressive sensing (CS) has attracted increasing attention in the areas of signal processing, computer vision and pattern recognition. In this paper, a new method based on the CS theory is presented for robust facial expression recognition. The CS theory is used to construct a sparse representation classifier (SRC). The effectiveness and robustness of the SRC method is investigated on clean and occluded facial expression images. Three typical facial features, i.e., the raw pixels, Gabor wavelets representation and local binary patterns (LBP), are extracted to evaluate the performance of the SRC method. Compared with the nearest neighbor (NN), linear support vector machines (SVM) and the nearest subspace (NS), experimental results on the popular Cohn-Kanade facial expression database demonstrate that the SRC method obtains better performance and stronger robustness to corruption and occlusion on robust facial expression recognition tasks.

Keywords: Gabor wavelets representation; compressive sensing; corruption and occlusion; facial expression recognition; local binary patterns; sparse representation.

Figure 1.

(a) The real part of the Gabor wavelet kernels at five scales and eight orientations; (b) The magnitude of the Gabor wavelet kernels at five scales.

Figure 2.

The process of LBP features extraction

Figure 3.

Examples of facial expression images from the Cohn-Kanade database.

Figure 4.

Recognition results of different methods with reduced dimension of Gabor wavelets representation.

Figure 5.

A corrupted image example (a) Original image of 640 × 490 pixels; (b) Resized image of 32 × 32 pixels; (c) 50% corrupted image.

Figure 6.

Recognition accuracy under different percentage corrupted.

Figure 7.

An occluded image example (a) Baboon; (b) Original image of 640 × 490 pixels; (c) Resized image of 32 × 32 pixels; (d) 30% occluded image.

Figure 8.

Recognition accuracy under different percentage occluded.