Unsupervised Detection of Apnea Using Commodity RFID Tags With a Recurrent Variational Autoencoder

• Main Authors: Chao Yang, Xuyu Wang, Shiwen Mao
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Apnea; deep learning; radio-frequency identification (RFID); recurrent variational autoencoder; respiration monitoring
• Online Access: https://ieeexplore.ieee.org/document/8720147/

With the rapid development of intelligent health sensing in the Internet of Things (IoT), vital sign monitoring (e.g., respiration) and abnormal respiration detection have attracted increasing attention. Considering the challenging and the cost of collecting labeled training data from patients with breathing related diseases, we develop the AutoTag system, an unsupervised recurrent variational autoencoder-based method for respiration rate estimation and abnormal breathing detection with off-the-shelf RFID tags. Moreover, for real-time breath monitoring, a novel method is proposed to cancel the distortion on measured phase values caused by channel hopping for FCC-complaint RFID systems. The efficacy of the proposed system is demonstrated by the extensive experiments conducted in two indoor environments, while the impact of various design and environmental factors is also evaluated.