Intro to Orbicularis Oculi

Orbicularis oculi is a muscle surrounding the eye. Though it is a single muscle, orbicularis can contract in distinct sections. The two relevant sections for this discussion are the:

• orbital area – or cheek raiser / AU 6 
• palpebral area – or lid tightener / AU 7

In FACS (Facial Action Coding System), the orbital area is referred to as cheek raiser, or AU 6; and the palpebral area is referred to as lid tightener, or AU 7. 

If you follow Paul Ekman-based emotion prototypes, cheek raiser is a key player for sadness and happiness (see “Duchenne smile”).** It is also observed in pain and common in many other expressions and occasions.

NOTE: **Recent post – “It’s All in the Eyes and Other Lies” breaks down the mainstream oversimplification on concepts like Duchenne smiles.

Lid tightener is one of the AUs listed in anger prototypes; however, it is by no means exclusively tied to anger. Lid tightener can also be used to express a variety of conditions such as sleepiness, concentration, frustration, trouble seeing, etc. 

NOTE: As a general rule of thumb, AUs in singular form do not necessarily define or represent an emotion. The face is complicated and meaning should be derived from additional contextual information. 

Posing Problems: What to Do When Your Model or Data Participant Cannot Perform the Target AU

Whether you’re using an expression model to pose AUs for character art references, machine learning data, or academic research, you will face challenges acquiring pure examples of cheek raiser and lid tightener. You will likely encounter traces of each AU in either attempted pose. Often times the model will contract lid tightener during cheek raiser. And in cases of high intensity lid tightener, there is a strong chance the model will also contract cheek raiser.

Another common mistake to be aware of when using expression models – specifically for cheek raiser – is the erroneous addition of the zygomaticus major, or lip corner puller / AU 12. Because people associate cheek raiser with happiness, and because cheek raiser is a difficult action to isolate, many people will unintentionally add a smile.

AUs represent the building blocks we use to communicate emotions, speech, and other valuable social information. Awareness of impurities in your AU references, regardless of what you may be observing them for, is essential.  For computer vision machine learning, awareness of AU impurities is crucial for data integrity, especially when considering emotion detection use cases. For art, understanding your AU references is integral for preserving the essence of human behavior, communication, and emotion. Even if you’re using stylized characters, you still must understand your references and how they behave in their original form.

There are some ways to mitigate the effects of co-activated AUs if you’re attempting to isolate cheek raiser or lid tightener:

• for character art:
  • If you are able to discern the distinct qualities of each action, you can subtract the unwanted effects of the co-activated AUs when you’re creating your 2D frames or blend shapes. Such qualities can be observed via changes in fat distribution, wrinkle depth, skin pull, etc.
• for computer vision machine learning:
  • If you’re using human labelers trained to identify distinct AUs as well as AUs within combinations, it will not matter if your participant data is impure. The labelers can classify cases of co-occurring data as multiple AUs, in this case – as both cheek raiser and lid tightener. Classifying contaminated data properly is necessary to avoid training your model to lump together distinct muscle actions. Lumping together distinct actions is bad data practice, as each action represents information used to convey emotion, attention, communication, etc.

WARNING:  While it’s possible to manually separate the effects of cheek raiser from the effects of lid tightener, if your model or participant is co-activating zygomaticus major, or lip corner puller / AU 12 – you should definitely put some extra time into training your model to mute their lower face. The push of fat distribution created by the lip pull will significantly change the look of cheek raiser, and it will not be as easy to differentiate as lid tightener and cheek raiser. Fortunately, it is much easier to get your model to stop smiling than to stop co-activating either orbicularis section.

References and Example GIFs

• cheek raiser 
• lid tightener
• cheek raiser + lid tightener combo

DISCLAIMER: I did what I could to isolate the expressions – but always be wary of action unit contamination. 

cheek raiser / AU 6 / orbicularis oculi, pars orbitalis

neutral

trace AU 6

max AU 6 shown in GIF

(1) neutral, (2) early trace of AU 6, (3) maximum intensity of 6 shown in this GIF 

When cheek raiser constricts:

• skin around the eyes moves in a pinching motion, tightening the outer corners of the eyes
• outer and upper areas of the cheeks are pulled up
• middle and inner parts of the lower eyelid may be slightly affected – but it is mostly the outer eye corners that tighten and pinch
• crow’s feet (wrinkles on the outer eye corners) may appear
• rings around the eye bunch up and tighten – which can pull down the outer corners of the brows
• eye bags may bunch and wrinkle

AUs 9 (nose wrinkler), 10 (upper lip raiser), and 12 (lip corner puller) also affect the cheek areas, but they affect the cheek areas in very different ways (to be discussed another time).

NOTE: The look of cheek raiser will change significantly with age and weight. In some people, the eye wrinkles may or may not be obvious. Older people often display a more wrinkly and pouchy look under the eyes. Children and people with more facial fat typically display less wrinkly and more pouchy eye bags.

Additional Cheek Raiser Examples

lid tightener / AU 7 / orbicularis oculi, pars palpebralis

neutral pose

trace AU 7

max intensity of 7 shown in GIF

(1) neutral, (2) early trace of AU 7, (3) maximum intensity of 7 shown in this GIF 

In lid tightener:

• skin under the lower eyelids tends to push up against the eyeball and toward the nose / tear duct
• movement is much different from that of cheek raiser – as cheek raiser focuses on the outer corners of the eyes and cheek area – and lid tightener does not
• movements are concentrated on the inner eye corners, narrowing the tear duct as the intensity grows

Observe the wrinkle pathway in the GIFs above and image below.

cheek raiser + lid tightener / AU 6 + 7 combo

GIF begins with cheek raiser alone, then lid tightener is added – creating a 6+7 combo.

trace AU 6

max AU 6 shown in GIF

max intensity of 6+7 displayed in GIF

(1) neutral, (2) early trace of AU 6 + 7, (3) maximum intensity of 6+7 shown in this GIF 

It’s easier to differentiate cheek raiser and lid tightener when you can see the transition from one to the other. The GIF above shows cheek raiser, but midway through lid tightener enters the equation, forming a combo shape.

The moment lid tightener arrives should be apparent. When lid tightener begins activating, the lower lid pushes up further in the medial area of the eye and all the way toward the tear duct, causing the tear duct to shrink significantly.

While cheek raiser can affect the look of the tear duct in extreme poses, the inner area of the eye is primarily affected by lid tightener.