Ecological Facial Display Recognition
The ecological theory of facial displays combined with the media equation
has implications for human-computer interaction, both for generation (of
the embodied agent's face), and for recognition (of the human user's face).
- Generation : An embodied agent's face must
- provide signals of future action . That is, it must
use its face to communicate something useful to the human user about what it is about to do.
- provide feedback that the user's facial displays are being received (and used).
- Recognition : An agent must
- use the fact that a human's facial displays will be correlated with the context of the
interaction , including the user's and the agent's actions.
- recognize facial displays adaptively.
That is, the agent must not make any a-priori
assumptions about what human facial displays are indicative of (including internal or emotional states).
The agent must adaptively learn how a human is using their face, and for what purposes. Clearly, the
most important facial displays to recognize are the ones which will help the agent perform its task.
These important facial displays will differ from application to application, and from user to user.
Therefore, a helpful agent will be one which learns categories of facial displays in an unsupervised
way. Read more about unsupervised recognition of facial displays
