Accelerometer-based motion analysis for elderly fall prevention and post-stroke rehabilitation

(eng) Fall in the elderly and stroke are nowadays both major public health issues inducing injuries, social isolation, impairments, huge costs to the society and even death. Therefore fall prevention and post-stroke rehabilitation assessment are research fields of interest. With the apparition of the Micro-Elecrtro-Mechanical-Systems (MEMS), small, light-weight, low-power and low-cost sensors could be developed to help medical practitioners in their diagnosis or to assess their treatments. This dissertation presents new accelerometer-based tools and related signal processing adapted for these challenges i.e: fall risk evaluation by elderly people and post-stroke rehabilitation assessment. A totally autonomous accelerometer network able to measure accelerations of the whole body was developed for fall risk assessment. This new tool extract movement features related to time, frequency, complexity and symmetry and uses classification algorithms to evaluate whether or not the subject presents a risk of falling. Some new features were computed, notably the correlation between both arm’s motion which seem to be one of the most risk assessment predictor. Frequency-related features have also shown good potential for fall prediction. In the field of post-stroke rehabilitation, the assessment of the rehabilitation process is usually made by clinical tests or visual-based motion tracking and kinematic analyses. These kinematic analyses provide an excellent, sensitive and objective tool. The major drawback of this kind of analyses is the use of high-cost visual motion tracking systems and their necessity for a dedicated room. Therefore a new feature extraction tool based on an accelerometer network and the Kalman filter has been developed and showed similar performances for kinematic analyses compared to those computed from the visual tracking systems. The major advantages of this new system are the portability and low-cost of the tool which allow its usage outside of the laboratory and even at home. A second system, using low-cost cameras and an accelerometer network has also been investigated for post-stroke rehabilitation assessment and 3D movement reconstruction. This system provides a 3D reconstruction solution robust against visual occlusions.