Using Wireless Tags to Monitor Bodily Oscillation

Traditional systems for monitoring and diagnosing patients' health conditions often require either dedicated medical devices or complicated system deployment, which incurs high cost. The networking research community has recently taken a different technical approach of building health-monitoring systems at relatively low cost based on wireless signals. However, the RF signals carry various types of noise and have time-varying properties that often defy the existing methods in more demanding conditions with other body movements, which makes it difficult to model and analyze the signals mathematically. In this paper, we design a novel wireless system using commercial off-the-shelf RFID readers and tags to provide a general and effective means of measuring bodily oscillation rates, such as the hand tremor rate of a patient with Parkinson's disease. Our system includes a series of noise-removal steps, targeting at noise from different sources. More importantly, it introduces two sliding window-based methods to deal with time-varying signal properties from channel dynamics and irregular body movement. The proposed system can measure bodily oscillation rates of multiple persons simultaneously, even when the individuals are moving. Extensive experiments show that our system can produce accurate measurement results with errors less than 0.3 oscillations per second when it is applied to monitor hand tremor.