Our face reveals many information to others. First, it uniquely identifies someone (except cases of twins of course). Secondly, it gives many clues about what someone think or feel, his/her attention (i.e. Gaze direction), age, gender and ethnicity . Facial expression analysis is therefore the study that is dedicated to analyze the facial expressions from an image/video of someone face. It’s about finding points in the face and relations between these points that can give the analyzer insights about what this face express now. Knowing and analyzing the facial expression is important for different aspects of human-to-human communication. Nowadays it becomes important too for human-machine interaction. Despite that you can think that face is not enough to show you what someone think and feel, many researches were done to predict how someone feel from his/her facial expression only.

The advances in technology and its adoption in our daily lives leads to an idea called human-machine interaction. Researchers and tech companies becomes interested to give the machines(software) which they build the ability to interact with human as natural as it can be. They found this task as challenging one, but they compete to reach the ultimate naturalness of human-machine interaction.

How Could Machines Know Your Facial Expressions?

For a machine, to analyze a face expression, it firstly should understand the facial clues from set of points from the face for each possible expression human makes. This stage is formally called feature extraction. We take the face image, extract points and relation to points that uniquely identify each expression. These features are not the same for all human faces. Faces are not identical. Therefore machine cannot just make a comparison, it should build a model. This model has the property that each facial expression can be represented by a range of features not only one. Therefore having new face that never seen before, machine can predict what expression is drawn on it. This model generated by machine learning algorithms. What define the effectiveness of machine to predict new expression, is who machine extract features and the how effective the model that discriminate features from each others. Also, feeding the learning algorithm by many different faces can make it more accurate in building the model.

The state of art for the methods  can be found in a variety of publications[2012]. You can also build your own facial expression analyzer using open source libraries that perform machine learning and image manipulation for you and free faces datasets.

Why it’s that Important?

  1. Human-Computer Interaction: as mentioned before, Enhancing the interaction, computer algorithms can be aided by knowing your reactions. This allow to build automated software that can adapt according to your reaction. This includes tracking your behavior so recommendation systems, digital marketing algorithms can use your data to be more accurate in finding the preferences and the target audience which benefit the users who use them.
  2. Human-Robot Interaction: Robotics and AI is the future. Industrial robots are already installed in factories and interact daily with human co-workers. Researches now is extremely focused in making the interaction between robot and human as natural as it can be. Making robot understand the co-worker reaction is essential to make a natural communication
  3. Research Purpose: Bio-metric data can be used to figure out human behaviors, helps in psychology researches and medical researches.

Facial Expression Analysis Products

As machine learning algorithms become more accurate and stronger, the art of doing facial expression analysis automatically become more stable area for industrial use. There are many companies that relies on facial expression analysis technology  to provide a set of products to their clients. Below is seven different companies which provide software, API and SDK solutions for facial expression analysis:

Affectiva provides an emotion recognition SDK and API which trained over million of images.

The company works in human behavior understanding using AI methods. One of parts to understand the human behavior is the face expression analysis.


FaceReader From  

A software to  a real-time facial expression analysis for images and video.

It provides bio-metric analysis (facial expression analysis is one of them) tools for research and industry use.

They provide API for facial expression analysis, the APIs function  can be demonstrated from their demo page.

nViso is 3D facial imaging product that extract emotions from 3D view of the face.

They provide SDK and API for real time facial expression analysis for images and video.

Leave a Reply

Your email address will not be published. Required fields are marked *