Fall Detection Algorithm Development

Project Description

Development of a new algorithm for human fall detection using JeVois camera. The main goal was to be able to operatively detect a fall of an elderly person. The algorithm was based on Hidden Markov Model to determine states and their transitions. Requirements for the algorithm: 1. Should be able to work with high FPS (60 and more). 2. Should be able to identify a person's state (sitting, walking, lying, falling) 3. Should be able to detect human inside the frame and detect fall of this person. 4. Should not detect lying as falling. Further development includes works on Regions of Interest for excluding these areas from analysis or taking more care when a person is in these areas. This may help to understand the everyday activity of a person and trigger specific events if some actions take more time than usual. List of technologies used: OpenCV, HMM, C++, Linux, MacOS, JeVois