Follow

Understanding the Facial Expressions Detection Suite

Upon click of the menu bar symbol, you will see the option for Detections. Clicking the arrow will allow for sub options to appear, including Facial Expressions. Clicking this will lead to the Facial Expressions Interface. 

On the left hand side of the interface is a simple avatar that will mimic your own facial expressions. (Note: the expressions are shown as if a camera view, not mirrored).

In the center of the panel is a series of graphs, indicating the various expression detection event signals. These graphs show a short history of the detections listed. The graphs can be interpreted as follows:

  • BLINK: low level indicates a non-blink state, while a high level indicates a blink.
  • RIGHT WINK / LEFT WINK: these two detections share a common graph line.  A center level indicates no wink, low level indicates a left wink and high level indicates a right wink.
  • LOOK RIGHT / LEFT: these two detections share a common graph line and a single sensitivity slider control.  A center level indicates eyes looking straight ahead, while a low level indicates eyes looking left, and a high level indicates eyes looking right.
  • RAISE BROW: low level indicates no expression has been detected, high level indicates a maximum level of expression detected.  The graph level will increase or decrease depending on the level of expression detected.
  • FURROW BROW: low level indicates no expression has been detected, high level indicates a maximum level of expression detected.  The graph level will increase or decrease depending on the level of expression detected.
  • SMILE: low level indicates no expression has been detected, high level indicates a maximum level of expression detected.  The graph level will increase or decrease depending on the level of expression detected.
  • CLENCH: low level indicates no expression has been detected, high level indicates a maximum level of expression detected.  The graph level will increase or decrease depending on the level of expression detected.
  • RIGHT SMIRK / LEFT SMIRK: these two detections share a common graph line.  A center level indicates no smirk, low level indicates a left smirk and high level indicates a right smirk.
  • LAUGH: low level indicates no expression has been detected, high level indicates a maximum level of expression detected.  The graph level will increase or decrease depending on the level of expression detected.

On the right hand side of the panel, two tabbed panels are present: Sensitivity and Training. These panels are explained further below. 

Sensitivity Adjustment Panel

The Xavier SDK offers sensitivity adjustments for the Facial Expressions detections. This is controlled through sliders to the right of the corresponding graph.

For each Facial Expression, check the performance of the detection. If you feel that the specific Facial Expression detection is not responding readily to a particular expression, then increase the sensitivity for that expression. If you feel that it is too easy to trigger a particular expression, or you are seeing "false positive" expressions, then decrease the sensitivity for that expression. Sensitivity can be increased or decreased by moving the sensitivity slider to the right or left, respectively.

The Facial Expressions detection suite supports two types of "signatures" that are used to classify input from the neuroheadset as indicating a particular expression. The icon to the right of the sliders is an indicator of whether the Universal Signature or Trained Signature is being used. A circle with three dots is shown when the Universal Signature is active, while a circle with one dot inside indicates that a Trained Signature is active. An empty circle indicates that a Trained Signature has been selected, but that no training data has been collected for this action, and so it is currently disabled. The default signature, the Universal Signature, is designed to work well for a large population of users for the supported Facial Expressions. 

If the application or user requires more accuracy or customization, then you may decide to use a Trained Signature, as is described below.  

Training Panel

In this mode, the Facial Expressions detection suite requires the user to train the system by performing the desired action before it can be detected. As the user supplies more training data, the accuracy of the detection of Facial Expressions typically improves. If you elect to use a Trained Signature, the system will only detect actions for which the user has supplied training data. The user must provide training data for a neutral expression and at least one other supported expression before the Trained Signature can be activated with the Use Trained Signature checkbox.

Important Note: Not all Facial Expressions can be trained. In particular, eye and eyelid related expressions (ex. "blink" or "wink") cannot be trained and always rely on the Universal Signature.

Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request

Comments

Powered by Zendesk