CSI-based respiration rate detection using commodity WiFi

Abstract

In recent years, with the popularity of WiFi, WiFi-based wireless sensing has become a research hotspot in academia. The detection of human activities and features has become a hot research topic. In the medical field, respiration frequency is of great value, but it still needs to rely on external sensors for sensing. Therefore, how to conduct contactless and efficient detection has significant research value. In this dissertation, we develope and evaluate two particular methods: CSI amplitude and phase difference-based tensor decomposition and CSI ratio-based independent component analysis. The tensor decomposition method shows exceptionally high accuracy but still requires improvements in computing efficiency. The CSI ratio-based method is unsatisfactory in accuracy and needs to be improved in the future. In addition, based on the above algorithms, we develop a real-time respiration sensing system, including a front-end visualized interactive interface. The system can receive streaming data and dynamically display the respiration frequency and respiration curve, reflecting specific application values.