Wearable Vision Detection of Environmental Fall Risks using Convolutional Neural Networks

A number of hazards in the home and public environment have been identified that contribute to falls and related injuries. These influences interact with intrinsic factors, such as poor vision or balance, to compound fall risk for seniors. A major challenge is a lack of available tools to assess individual risk, particularly the frequency of exposure to specific hazards. For example, an individual with unstable gait may be at greater risk of falls with frequent exposure to hazards than another with similar capabilities but little or no exposure. The aim of the current study is to examine the potential of wearable egocentric cameras (e.g., GoPro), coupled with advances in machine learning techniques, to detect fall risk hazards.