Sites Inria

Version française

From biometrics to health care

Stéphanie Salti - 15/12/2017

Analysing facial expressions: from biometrics to health care

Antitza Dantcheva (Stars) Antitza Dantcheva (Stars)

Antitza Dantcheva, researcher in the Stars team at the Inria Sophia Antipolis - Méditerranée centre, is a recipient of the ANR JCJC (Young researchers grant). She is currently working on concrete applications ranging from security to health care.

Face and emotions are unique

Reading faces is within the reach of almost everyone: looking at a face, it is possible to deduce identity, gender, but also age, ethnic origin or the intelligence of a person. "Now, thanks to recent advanced algorithms, such as deep learning, we can go beyond this simple analysis by reading more features," explains Antitza Dantcheva, recipient of the ANR JCJC grant and member of the Stars team. "Exploring additional features in a face allows us, through computer vision algorithms, to identify mental or psychological states."
For the past nine years, Antitza Dantcheva has specialized in the field of facial analysis: after completing her doctoral thesis at Eurecom and Telecom ParisTech in 2011, she was a postdoctoral fellow at the American universities of Michigan State and West Virginia – the two biggest powerhouses of research in biometrics - before joining Inria in 2014 with the Labex scholarship and the prestigious Marie Curie Fellowshop.
As part of her recently launched JCJC research project "Envision: Computer Vision for Automatic holistic analysis of Humans", the researcher, assisted by a Ph.D. student and a post doctoral fellow, aims to develop reliable methods of facial analysis and to study more advanced mental and psychological states. The aim is also to demonstrate how unique a face and its emotions are. The project is particularly innovative, and it targets analyzing data on a large scale.

Detecting possible terrorist attacks

This type of fundamental work naturally has a variety of applications. One application of this work is related to security in airports or public places. "After the recent terrorist attacks, Europe has begun a more in-depth debate on the use of multimodal and large-scale biometrics for the protection and security of citizens," says Antitza. "Thanks to a thorough analysis of the face, our research could enable us to detect possible terrorist attacks in the distant future. We could identify recurrences of emotions or behaviors in a number of subjects in crowded environments."

Helping Alzheimer's patients

Another domain of application concerns the health sector. As part of her participation in CobTek, a host team from Nice Sophia Antipolis University (see box), Dr. Dantcheva paid particular attention to Alzheimer's disease patients. "The recognition of expressions in these patients is indeed essential as they have lost much of their cognitive ability and sometimes their ability to communicate verbally."
If this process is unavoidable, solutions can be developed to delay the onset of symptoms by promoting a more active life for patients. This is the subject of another work by Antitza Dantcheva. She uses facial expression recognition in music therapy: "The objective is to develop algorithms to objectively identify the music that resonates with these patients. From then on, these people are able to relive the memories associated with this music completely and, free of their anxiety, become more independent and able to perform acts of daily life again." The study currently concerns approximately 200 patients. But expanding the number of data could considerably expand the scope of this research.
"This project could reduce the financial costs associated with the disease," says Antitza Dantcheva. But above all, it is about promoting independent patient management and self-management." At the beginning of the decade, some 35 million people worldwide were suffering from dementia.

CoBTek places news technologies at the service of elderly patients 

Created in January 2012, CoBtek (Cognition - Behaviour - Technology) is a welcoming team from the University of Nice Sophia Antipolis. One of its objectives is to develop new knowledge about elderly patients using new technologies. In collaboration with the CMMR of Nice (Centre Mémoire de Ressources et de Recherche), the CobTek team conducted an experiment on patients with Alzheimer's disease: "It is difficult for a doctor to realize the real state of an Alzheimer's patient since the subject suffers from more or less severe cognitive disorders", explains François Brémond, research director at Inria. "However, a precise diagnosis can enable a practitioner to make the right decision about the best supervision: a foster home or even a home."
Thanks to new technologies, the CobTek team has developed a cognitive pathway test to replace the traditional, subjective questionnaire. "Faster, this questionnaire is playful, but it also validates certain cognitive and motor functions," explains the research director. Drawing on the resources of computer video, the CobTek team has also developed new algorithms to analyze patient videos in different situations: "From these videos, we will be able to measure the fragility of a person's walking or even the way he or she performs a specific task of daily life, and especially if he or she manages to perform this task in accordance with the instructions given," continues François Brémond. For the physician, these data allow for a much more accurate and earlier diagnosis: "Based on all the data collected, we can also produce statistics on all the patients who consult the Resource and Research Memory Centre in order to improve their care and quality of life".