An intelligent vision-based system for detecting and classifying human behaviors

Human action recognition is an important computer vision field with applications ranging from video surveillance, robotics, to automated driving systems among others. We are mainly interested in the constant monitoring and the rapid detection of self-harm events. Our interest is motivated by the protection of inmates in correctional settings. During the last decades, several technologies have been developed for preventing self-harm attempts in prisons. However, most of them use cumbersome devices and/or are greatly depending on human guard attention and intervention.

Our objective is to develop autonomous video surveillance solutions in order to intelligently detect life-threatening events and actions, and allow a rapid and reactive intervention. These objectives will be achieved by ensuring the efficient exploitation of recent advances in visual acquisition technologies, computer vision methods and artificial intelligence.

We plan to develop computer vision methods for gait analysis and fall detection. We plan to use the combinations of several streams provided by RGB-D cameras. We will base our work on existent methods that are mostly developed in the context of neurodegenerative diseases and elderly people monitoring. We plan to extract and measure human gait features such as a center of gravity change, a pace length variation, and a joint angle variation. We plan to use these features for constructing a motion model in order to detect unsteady walking and falls. Approaches based on machine learning will be used along with existing datasets.
Our work will allow our industrial partner to develop a commercial product to be deployed in correctional services. The impact of this product is expected to help saving the lives of incarcerated people at risk, and avoid the resulting suffering.
 

Connaissances requises

  • Excellent knowledge in artificial intelligence algorithms
  • Excellent motivation
  • Good research record and/or good academic curriculum
  • Ability to work as a team and independently
  • Excellent communication skills especially in writing
  • Excellent programming knowledge
     

Programme d'études visé

Doctorat

Domaines de recherche

Technologies de l'information et des communications

Financement

Bourse du professeure

Autres informations

Date de début : Aussitôt que possible

Personne à contacter

Rita Noumeir | rita.noumeir@etsmtl.ca