Radio-frequency (RF) sensing for deep awareness of human physical status

Abstract

In recent years, there are a rise in studies of contactless Radio Frequency (RF) sensing for human physical status like one’s respiration behaviour. In these studies, a radar sensor will be used to collect the raw data from a human subject to provide insights on the respiration. This conventional method includes the use of complex signal processing and domain expert for feature engineering to produce a result. This came the motivation of using deep learning to leverage or avoid this complication. In this project, an attempt is made to implement deep learning techniques on relatively unexplored field of predicting human respiration rate. Techniques like Denoising Convolutional Autoencoder (DCAE) is applied to denoise the potential noisy data then coupled with a 1-Dimensional Convolutional Neural Network (1-D CNN) to learn from these processed data to make a prediction. Despite the prediction results shown in this project are far from ideal, the proposed approach can be a good foundation for future study to build on.