Facial expressions convey non-verbal cues, which play an important role in interpersonal relations. automatic recognition of facial expressions can be an important component of natural human-machine interfaces; it may also be used in behavioural science and in clinical practice.
Although humans recognise facial expressions virtually without effort or delay, reliable expression recognition by machine is still a challenge. This paper presents a high-level overview of automatic expression recognition; it highlights the main system components and some research challenges.
MAIN ARCHITECTURAL COMPONENTS:
A. Generic Architecture:
Despite the task duality that exists between facial expression recognition and face recognition, it can be observed in the literature that similar architectures and processing techniques are often used for both recognition tasks. The duality arises from the following considerations. In addition to conveying expressions, faces also carry other information such as the identity of a person. By definition, the expression of a face is the facial element in facial expression recognition.
B. General Description:
With regard to the interconnection between the blocks shown in Figure 1, feedback paths between blocks are absent from most expression recognition systems, although feedback could be beneficial for improving recognition accuracy.
Although humans recognise facial expressions virtually without effort or delay, reliable expression recognition by machine is still a challenge. The problems that have haunted the pattern recognition community at large still require attention.
The controlled imaging conditions typically cover the following aspects:
- View or Pose of the Head. Although constraints are often imposed on the position and orientation of the head relative to the camera, and the setting of camera zoom, it should be noted that some processing techniques have been developed, which have good insensitivity to translation, scaling, and inplane rotation of the head.
- Environment Clutter and Illumination. Complex image background pattern, occlusion, and uncontrolled lighting have a potentially negative effect on recognition. These factors would typically make image segmentation more difficult to perform reliably.
- Miscellaneous Sources of Facial Variability. Facial characteristics display a high degree of variability due to a number of factors, such as: differences across people (arising from age, illness, gender, or race, for example), growth or shaving of beards or facial hair, make-up, blending of several expressions, and superposition of speech-related (articulatory) facial deformation onto affective deformation.
This paper has briefly overviewed automatic expression recognition. Similar architectures and processing techniques are often used for facial expression recognition and face recognition, despite the duality that exists between these recognition tasks. 2-D monochrome facial image sequences are the most popular type of pictures used for automatic expression recognition.
Although a variety of face detection techniques have been developed, robust detection and location of faces or their constituents is difficult to attain in many cases. Features for automatic expression recognition aim to capture static or dynamic facial information specific to individual expressions. Geometric, kinetic, and statistical or spectral-transform-based features are often used as alternative representation of the facial expression prior to classification.
A wide range of classifiers, covering parametric as well as non-parametric techniques, has been applied to automatic expression recognition. Generally speaking, automatic expression recognition is a difficult task, which is afflicted by the usual difficulties faced in pattern recognition and computer vision research circles, coupled with face specific problems.
As such, research into automatic expression recognition has been characterised by partial successes, achieved at the expense of constraining the imaging conditions, in many cases. Unresolved research issues are encapsulated in the challenge of achieving optimal pre-processing, feature extraction or selection, and classification, under conditions of data variability.
Sensitivity of automatic expression recognition to data variability is one of the key factors that have curtailed the spread of expression recognisers in the real world. However, few studies have systematically investigated robustness of automatic expression recognition under adverse conditions.
Source: Staffordshire University
Authors: Claude C. Chibelushi | Fabrice Bourel