Non-contact respiratory rate estimation based on RF signals with machine learning algorithms

Respiratory rate is a very important parameter in the field of health monitoring. In any clinical or emergency medical situation, monitoring the patient's respiratory rate is one of the one of the critical first examinations to understand the general condition of the patient and to determine the intervention to be performed. Researchers continue to work on developing a highly accurate respiratory rate estimation method within limited observation period. In this study, we propose a new non-contact respiratory rate estimation method using radio-frequency (RF) signals and machine learning models within a limited observation time. 10 different people's respiratory data are used in this study, and software-based radios are used to generate and receive RF signals. The performances of various machine learning models on the dataset are extracted and the model that gives the best result is used. Experimental results show that the proposed method achieves an error rate of 2.33% for a limited observation time of 10 seconds and is superior to the existing methods in the literature.