Device-free detection and localization of people using UWB networks

Ultra-wideband (UWB) transmission is a promising technology for indoor localization, due to its high-resolution ranging and obstacle penetration capabilities. Most practical UWB localization systems rely on targets (e.g., objects, people) to carry an active UWB device. In some scenarios (e.g., intruder detection, elderly care, smart environments, emergency response) it is desirable to have the ability to track people and assets in a passive manner, without requiring them to be equipped with any radio-frequency (RF) device. The high time resolution property of UWB radios can make it possible to resolve the human-body reflected path among all the other multipath reflections. This reflected path changes over time due to minuscule movements, even though the person seemingly stands still. Because of this, the corresponding received signal samples also vary over time. This observation is presented for an experimental UWB system and forms the basis of the device-free person detection and localization technique in this thesis. The technique collects the energy in the variations of the signal and estimates the travelling distance for the human-body reflected path through its delay estimation. Each estimate draws an ellipse around the transmitter and the receiver for the position of the person. By combining multiple ellipses, a unique position estimate is given for many different positions of the person in an indoor environment. The received signal samples show different variations depending on the random movement of each person. This gives us a way to detect a second person in the environment without extending the measurement setup. By quantifying the correlations between the samples of the received signal, we can understand if the samples are affected by the same person or not. The correlation value is higher if samples are affected by the same person and remains low if the samples are affected by different persons. A detection and device-free ranging method is given for the second person. A localization method for multiple persons is also developed by combining the multiple link correlations