2020 – 2021 – Wireless Sensing, Machine Learning, Signal Processing
Research Engineering (Java, Python, PyTorch)
Vital Sign Monitoring and Movement Tracking using Commodity-Off-The-Shelf Channel State Information
This report describes the results and implementations of my research project on using Channel State Information (CSI) obtained from Commodity-Off-The-Shelf devices for vital sign monitoring and movement tracking as well as activity detection. It describes how it is possible to obtain CSI from different types of devices and investigates their respective properties, advantages, and disadvan-
tages. Additionally, ground truth on respiration (chest acceleration) and heart rate (ECG) is obtained for verification. Different approaches for activity, respiratory, and heart rate detection with a focus on heart rate variability using the provided CSI are discussed, implemented, and evaluated. Afterward, possible approaches for increasing the reliability of the detection systems are described.
Combining amplitude and phase information using the complex conjugate product representation seems promising and processing algorithms for sanitizing and preparing it for further analysis are developed. Afterward, state and movement classification on the obtained data is performed using machine learning techniques. The framework for obtaining, collecting, recording, processing, and replaying CSI and other types of data in real-time that was developed as part of this project is presented and released publicly.
