Facial Emotion Recognition
Overview
One of the tasks in human machine interaction is emotion recognition. Facial Emotion Recognition (FER) is the ability to interpret facial expressions captured via an image acquisition system into a corresponding emotion class by implementing machine learning techniques.
This system can be useful in rehabilitation therapy by helping to treat patients suffering from Autism Spectrum Disorder (ASD).
A virtual world tailored to the individual, combined with an educative gaming platform which rewards emotional interactions creates the appropriate atmosphere for the treatment and provides a positive outlook towards ASD patients' interactions and behavioral development.
Current research at ITeM
In this work, images of subjects expressing different emotions, gathered from multiple datasets, are analyzed and classification models are constructed.
The main topics are:
- Different neural network models are trained tested to obtain a robust model
- Predictions are analyzed to validate the neural networks decisions
- Developing a closed loop system to provide real-time predictions of the emotions
- Improving robustness of neural networks through different approaches