• PhD Student:
  • Sara Kebir
  • Funding : ANR, Artois
  • PhD defended on :
  • Dec 20, 2023

With the extension of the life span, the aging of the French population continues. People aged 65 and over represent 20% of the population. INSEE’s demographic forecasts predict that the proportion of people over 60 years of age will increase strongly until 2035 and estimates that 4 million seniors would be in loss of autonomy in 2050. This problem affects all of France, but it has certain specificities in Hauts-de-France region, such as the fact that seniors in the region are more vulnerable to poverty than at the national level. There is then a real question of society regarding the care of this segment of the population. One of the solutions is of a technological nature and concerns Smart homes specially designed for the elderly.

Smart home technologies can bring real health and wellness benefits in addition to energy savings, comfort and safety. Elderly solutions also pose other constraints because of their state of health and, possibly, their loss of autonomy, loneliness for some of them and their relationship to technology. The existing offer in Smart home technologies mainly concerns home automation in the broad sense and is hardly adapted to the needs of the elderly. In addition, most existing solutions are expensive, complex to deploy, and often require controlled and constrained environments. It is clear that lightweight, resource-efficient, inexpensive, scalable solutions are required, which requires minimal effort of installation, configuration and control by users. It is in this perspective that the Cluster Senior University strives to develop projects and take initiatives to propose effective solutions to promote home care by meeting the specific needs of seniors. The thesis topic proposed here is in computer science but contains a component which falls within the social sciences. Indeed, the study of real needs and the acceptability and usability of a Smart home solution by seniors is essential. This component will be conducted in a sociological perspective, and will allow to analyze quantitatively and qualitatively data from surveys conducted in the past.

The design, prototyping and experimentation of a Smart home solution adapted to the constraints and needs identified will be the major component of this thesis. It will be a question of designing and validating a prototype solution for the Smart home specially adapted for the elderly. This prototype will be provided with means of:

  • Data collection from a set of sensors (eg motion sensors, temperature, brightness, etc).
  • Recognition of activities to predict from the observed data (and possibly past activities) what the older person is doing as an activity (eg sleeping, eating, phoning, reading, cooking, etc.). This is typically a problem of annotation of sensor data streams.
  • Detection of abnormalities and risky situations: It will be necessary to detect rare, exceptional events or significant deviations from a normal behavior supposed to be known (eg awakening too late, absence/excess in certain activities). We can also consider explaining the anomalies in a perspective of recommendation or decision making (eg heating, ventilation, air conditioning).

A lightweight solution is needed for cost and power considerations in addition to the fact that it is necessary to store and process the data locally for privacy considerations. Major challenges will be to adapt some existing approaches to run locally will very limited ressources and use only the resident data. Another big challenge is personalization and taking into account needs and preferences of each user. The research topics of this thesis mainly comes from artificial intelligence with an application in social sciences. The thesis supervision team is complementary and already has experience in the relevant fields.