Facial recognition and facial expression analysis both examine people’s faces; the latter gives insights into what emotional state a person may be in. Application-wise, use cases range from security to recruitment to business intelligence.
Facial recognition and facial expression analysis both examine people’s faces; the latter gives insights into what emotional state a person may be in. Application-wise, use cases range from security to recruitment to business intelligence.
As humans, we go through a range of emotions every day. Seeing an old friend that you haven’t seen for a while and having a fight with someone you love trigger different emotions. Indeed, having emotions is one of the fundamental characteristics that make us human.
Perhaps the best indicator of one’s emotional state is facial expression. Whether it’s raising the eyebrows, opening up the eyes or pursing the lip, one’s emotion can easily be detected through facial expressions. That said, facial expression analytics, which helps identify a person’s possible emotional state by analyzing what’s showing on or through his face, can be an effective tool helping end users achieve their specific objectives.
“The automatic analysis of facial expressions is motivated by the essential role that the face plays in our emotional and social life. Facial expression is one of the most compelling and natural means that human beings have to communicate our emotions, intentions, clarify and emphasize what we say, as well as to indicate understanding and disagreement,” said Laura Blanc Pedregal, CMO of
Herta. “Furthermore, unlike other non-verbal channels, facial expressions are cross-cultural and universal, not depending on the ethnicity, age and gender of the individual.”
Different applications
Indeed, with the aid of computer vision and AI, expression analytics have become more mature and accurate. Demand is set to grow as well.
According to Gartner, by 2022, 10 percent of personal devices will have emotion AI capabilities.
In terms of applications, facial expression analytics can be employed in a range of use cases that include the following:
Security and law enforcement: The analytics can identify someone who is potentially violent, making it ideal for a variety of verticals for example enterprise and education. In law enforcement, facial expression analytics can come in handy as well. “In the law enforcement sector, it can be used for questioning suspects. It is interesting for the detection of emotional incongruities, that is, situations in which the subject verbally expresses an emotion while showing a very different one on the face,” Pedregal said.
Recruitment: Human resources can use the technology to help determine whether someone is fit for a position, especially those that are critical in nature. “It can help build psychological profiles of relevant employees like dangerous goods drivers, high profile executives, security officers or air traffic controllers. In the education sector it is useful to know the people who deal with children,” Pedregal said.
Business intelligence and others: In terms of business intelligence, facial expression analysis can also provide value. Retailers, for example, can use the technology to decide the general feelings of customers visiting their store and make the necessary adjustments accordingly. In the
smart home arena, further automation can be achieved – if the smart home system decides the user is angry, for example, mellow music will be played.
Certain challenges
Certain challenges associated with facial expression analytics still remain.
Privacy, for example, can be an issue, and user consent may be required and all data should properly kept. The end user entity should also find out whether there are laws or regulations applied to the use of
facial/emotion recognition technologies in their particular region.
Then, there are also concerns that there are certain instances where facial expression analytics may be less accurate or work less well. “(Herta’s BioObserver) has been trained and validated with relevant facial microexpression databases, which are quite balanced in terms of demographic groups. They are balanced in terms of gender and ethnicity. So the bias in that sense should not exist. However, in this case I would say that the age of the person is more key: wrinkles due to age can be more challenging than ethnicity and gender in this sense,” Pedregal said.
According to her, the BioObserver is able to detect basic facial emotions such as "joy,” “sadness" or "anger” and more subtle micro expressions of the face such as "frown," "blink" or "eyebrows raise.” She adds that BioObserver, based on deep learning, also allows extracting the direction of the gaze and the orientation of the head to monitor behavioral metrics such as the degree of attention of the individual.
“BioObserver analyzes the face frame by frame, either from a pre-recorded video or from a camera capture in real time. It works on the CPU (both video and live analysis) so it does not require a lot of computer power. Videos can be analyzed at a speed faster than real-time so it saves a lot of time to the user compared to current techniques which process the video at a real-time speed,” Pedregal said.