Human action recognition
Finding and identifying human actions from real time data, in particular video data, is an important challenge, with a variety of potential applications ranging from medical applications such as monitoring rehabilitation of patients, over sports to safety in the workplace or in public places. In many of these areas it is crucial to preserve the privacy of the person who’s actions are being monitored. One effective approach to achieve this, taken by the research group, is to extract matchstick figures as seen in the image to represent human pose in a de-identified way.
The annual cost of work-related injuries in the UK is estimated at £4.6bn, and these injuries are often due to poor practice and training in manual handling. The Health and Safety Executive (HSE) research how to improve safety in the workplace in the UK, often using manual frame-by-frame analysis of video data at a substantial cost. Through the Alan Turing Institute the research group have linked up with HSE to use signatures and computer vision to automate some of this process and contribute to making public places safer. The group is working to further their abilities to automatically understand human activity and explore more of the potential impactful applications.