The Full Story
Autonomous Senior Helper System for Enhanced Safety and Well-Being
In today’s society, the proportion of seniors within the population has increased significantly due to advances in healthcare and the decline in the birth rate. According to research conducted by the Pew Research Center, about 27% of adults ages 60 and older live alone in America. With no one to take care of them, seniors who live alone are exposed to many risks such as falls, heart attacks, and overlooked symptoms or illnesses. For seniors who live at home by themselves, it is quite difficult for them to seek help when such incidents happen. Therefore, our goal is to develop a light-weighted computer vision system that can monitor the activities of old adults through cameras and raises an alert to notify their family members, the emergency department, and any other related parties when seniors being monitored fall or have a heart attack, so that they can be helped as soon as possible. We aim to build a reliable system that can be easily deployed and does not rely on advanced hardware like GPUs. It should serve as a guardian that can look out for seniors 24/7. Despite all the great functionalities, it should still be affordable to most seniors in need.
Our Focus
We currently focus on developing a Computer-Vision-based fall detection system and can detect human falls regardless of any occlusions in the scene. More specifically, we are working on a generic action recognition algorithm with application on fall detection. This algorithm should be easy to generalize to detect other medical emergencies like heart attacks.
Results
We have successfully developed a skeleton-based fall detection system using Spatial-Temporal Neural Networks (STGCN). We used AlphaPose to extract skeleton data from each video frame and STGCN for inference. More details can be found in our technical report and GitHub repository.
