의료형 IoT 센서 통합 인공지능 기반 엣지 시스템 : 스트레스 모니터링

Abstract

Mental stress has widespread implications for both individual well-being and society. As bio-signals are increasingly used for stress measurement, the management of large data volumes led to the necessity of a cloud system for Artificial Intelligence (AI) computations. Due to latency and privacy issues associated with using a cloud system, we propose an AI-computable edge system embedded onto a wearable medical sensor for stress assessment. In this study, we designed an abdominal wearable sensor capable of collecting electrocardiogram, electrogastrogram, and respiratory waveform. The sensor incorporates a deep neural network (DNN)-based model for stress detection and triggers alerts when high level of stress is detected. The DNN-based stress detection model we developed achieved an average accuracy of 88.8%, sensitivity of 88.7%, specificity of 89.0%, and F1-score of 0.885. As future works, we will evaluate the feasibility of our stress-monitoring edge system through human subject experiment.