Mining Wi-Fi Channel State Information for breathing and heart rate classification

Current vital signs monitoring systems require that the subject wears sensing devices. An alternative approach is using device-free technologies such as the Channel State Information (CSI) of a Wi-Fi signal. However, recent works using CSI for vital signs monitoring rely on complex signal processing techniques to improve its reliability. Considering that breathing and heart rate provide relevant information about the current health status of a subject, in this paper we develop an experimental system that combines signal processing techniques, such as filters and time and frequency domain analysis, with Data Mining techniques for breathing and heart rate monitoring. We also provide a thorough analysis for understanding CSI data as a technology for vital signs monitoring. Using K-Nearest Neighbors, Support Vector Machines, and Quadratic Discriminant Classifier models, our system achieves an accuracy of 99.18% for breathing rate classification while identifying heart rate monitoring challenges that are also stated in this paper.