Deep learning approaches for facial expression recognition

Recent studies point to the increasing importance of facial expression recognition (FER) in different application areas like healthcare and assistive technologies. Among these, computational diagnosis and evaluation of mental or facial diseases, machine-assisted rehabilitation, clinical psychology and psychiatry, pain monitoring, can be highlighted. FER also plays a relevant role in the case of diagnosis or assessment of cognitive impairments, such as autism and schizophrenia. The literature shows that FER systems effectively working in a health setting is still an open research problem. This paper presents a study on techniques for FER based on images, applying deep learning algorithms. A comparative analysis is performed, pointing out the combinations of algorithms and repositories that manage to obtain gains in the accuracy index.