0478 Personalized Exercise Intervention Delivery for People with Insomnia

Abstract Introduction Difficulty falling asleep, decreased sleep quality, and impaired daytime function are common in people with insomnia. Studies support benefits of appropriate exercises in people with insomnia, such as Tai Chi, an exercise that involves mind-body movement, meditation, and breathing techniques. Such multimodal exercise has been shown to be effective as an intervention to improve sleep, promote mental and physical health in clinical trials. Pre-recorded instructional videos are commonly used in these trials, but people with insomnia are not naturally motivated to exercise following the videos. Therefore, we aim to develop technology-enhanced platforms to facilitate their practice. Methods Previously, we successfully developed an IoT-based, home-oriented medical system, Medcare4Home, to support health-record tracking. Recently, we have designed and implemented a new feature to support and engage people with insomnia to practice Tai Chi for better sleep. This feature allows users to choose settings that are appealing to them, and to customize Tai Chi demo videos based on their own preferences. This new feature allows dynamic and real-time video background changes within various themes, and it also enables anytime audio setting changes (e.g., original soundtrack, light music, natural sounds, complete mute, etc.). We have investigated different state-of-the-art multimedia technologies to realize the idea, including video processing (OpenCV, MediaPipe, MODNet, RVM, MatAnyone) and audio manipulation (FFmpeg). Results Integration of MediaPipe and OpenCV performed better than other techniques. The Medcare4Home system was successfully converted to Amazon EC2, and the proposed new feature was integrated by MediaPipe and OpenCV, which allows customizable settings for video-guided exercises. Our evaluation testing showed that the system can load videos and adapt real-time choice of settings smoothly with only minor time delay. The system also allows researchers or healthcare providers to track users’ exercise adherence remotely. These new features allow Medcare4Home to deliver personalized exercise interventions in home settings. Conclusion Existing technologies enable new features in systems like ours to be implemented in home settings to enable multiple users to access personalized exercise for better sleep. Our next steps include efforts to minimize time delay, to include more evidence-based exercises, and to recruit patients with insomnia to test the system for refinement. Support (if any)