Research On Device-free Breathing Detection Via Channel State Information

Nowadays,smart life is a research hotspot.With diversified sensors,human beings continue to perceive themselves as well as the environment in a more efficient and accurate way.With the continuous integration of smart devices into people’s lives,they have played an increasingly important role in livability,safety,and health.A wonderful era is beginning.Breathing detection via wireless signals has been proved to be feasible,but there are still some problems that should be addressed.For example,current solutions usually require a large space.Also,they are vulnerable to various environmental factors such as surrounding people’s activities.Therefore,this dissertation first mathematically analyzes and compares the radius of the Fresnel zone in different directions.Then,an optimization model called WI-BD has been built to guide the antenna setting exploring the classic path-loss model in free space and the physical characteristics of common WIFI antennas.We also point out the relationship between the antennas distance and the effective detection range combining both theoretic analysis and empirical studies.We then propose IRFE,an efficient algorithm to extract the instantaneous breathing frequency.We effectively improve the recognition accuracy of power spectral density algorithm by polynomial fitting.The experimental results have confirmed the efficiency of our WI-BD and IRFE algorithms.The main contributions of this paper are as follows:1.We confirm the x direction as the best direction,and it nearly achieves the best effect of breathing detection based on the WIFI signal in theory.2.Combining both theoretic analysis and empirical studies,we quantify the correlation between the antenna spacing and the effective detection range,which provides a theoretical basis for the related experimental settings.We also point out that the miniaturization detection model can be feasible.3.We propose the WI-BD model and IRFE algorithm,and effectively improve the recognition accuracy of power spectral density algorithm.Specific solutions are designed for the three classic sleeping postures,especially the prone position is very difficult to be detected.