NSF Award
2401928
A trustworthy alert system that utilizes artificial intelligence on real-time video can significantly enhance response times to emergencies such as heart attacks or sudden falls. This technology necessitates trustworthy and accurate inferences of continuous streams of high-resolution video from multiple cameras. This project entails overcoming three challenges: 1) training AI models on potentially biased datasets that may lead to inaccurate detections; 2) maintaining HIPAA-compliant privacy while avoiding unintelligent image distortions that can impact inference accuracy and processing time; and 3) addressing the trade-off between latency improvement and inference accuracy when reducing image quality to optimize processing time for deep neural networks. <br/><br/>This project, a partnership between the School of Applied Computational Sciences, Meharry Medical College, and Hunter College, City University of New York, addresses those challenges through a trustworthy, privacy-preserving edge-native framework that supports a latency-sensitive healthcare response application for video. Its aim is to develop a novel semantic-expression-based incident detection approach that identifies and isolates privacy-sensitive scenes within the video frames. An important focus is to design and develop a deep neural network partitioning approach to real-time video analysis that is latency aware and optimized for data privacy. This technology will generate a trustworthy emergency alert and a report, explaining the human action that prompted the alert, that can be understood by both the healthcare provider and the patient. Discoveries will foster new research pathways that lead to fundamental advances in trustworthy AI, privacy-preserving video, and image processing in the healthcare domain. Moreover, this project includes outreach activities such as organizing special sessions and workshops at top-tier conferences, designing new courses and modules, and providing research experiences to students from underrepresented communities.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
