Fall Detection and Activity Recognition Using Human Skeleton Features

摘要: Human activity recognition has attracted the attention of researchers around world. This is an interesting problem that can be addressed in different ways. Many approaches have been presented during last years. These applications present solutions to recognize kinds activities such as if person walking, running, jumping, jogging, or falling, among others. Amongst all these activities, fall detection special importance because it a common dangerous event for people ages with more negative impact on elderly population. Usually, use sensors detect sudden changes movement person. embedded smartphones, necklaces, smart wristbands make them “wearable” devices. The main inconvenience devices placed subjects’ bodies. might uncomfortable and not always feasible this type sensor must monitored constantly, used open spaces unknown people. In way, from video camera images presents some advantages over wearable sensor-based approaches. paper vision-based approach recognition. contribution proposed method falls only by using standard video-camera without need environmental sensors. It carries out human skeleton estimation features extraction. opens possibility detecting but also kind several subjects same scene. So real environments, where large number may at time. evaluated UP-FALL public dataset surpasses performance other systems dataset.