In Western countries, the growing population of seniors brings us to think about new healthcare systems to ensure the safety of elderly people at home. Falls are one of the greatest risk for seniors living alone. Computer vision provides a promising solution to analyze people behavior and detect some unusual events like falls. In this paper, we propose to detect falls by analyzing the human shape deformation during the video sequence. The elderly silhouette is tracked along the video sequence using shape context matching. These silhouettes are then used to quantify the deformation based on Procrustes shape analysis. Our algorithm gives very promising results on video sequences of daily activities and simulated falls.