UC Santa Barbara
Ambient Intelligence Using Wireless Signals
- Author(s): Depatla, Saandeep
- Advisor(s): Mostofi, Yasamin
- et al.
With a recent increase in the number of wireless devices around us, there is a great interest in using wireless signals to sense and understand our surroundings. Wireless sensing enables several applications and presents us with unique opportunities such as sensing behind walls and preserving the privacy of humans involved. Therefore, there has been a steady growth in the research interests in this area in the recent past. In this dissertation, we focus on utilizing off-the-shelf wireless devices and fundamentally understand the information carried by the wireless signals about the surroundings.
This dissertation is focused on passive wireless sensing using off-the-shelf wireless devices. Since most off-the-shelf wireless device can only make basic measurements such as the power of the signal, the focus of this dissertation is to enable wireless sensing using minimal power measurements from the devices. Furthermore, to preserve the privacy of human subjects in the area, we focus on passive wireless sensing, i.e., without depending on people to carry any device. Moreover, as we may not have a priori access to the area of interest, we develop frameworks for sensing that minimize the requirement of prior calibrations.
This thesis then contributes to the area of wireless sensing through three main topics
1) Robotic through-wall imaging,
2) Occupancy estimation, and
3) Joint crowd counting and crowd speed estimation.
First, in Robotic through-wall imaging, we utilize unmanned ground robots with standard WiFi connectivity to enable a high-resolution imaging of an area behind walls. We use theories from electromagnetic literature to mathematically characterize the signal propagation, sparse signal processing for efficient processing, and proper path planning of robots to enable high-resolution imaging of the area. We image several structures using our framework and present the experimental results.
Next, in occupancy estimation, we first show how to count the number of people walking in an area using only a single standard WiFi link and without relying on people to carry a device. Through a new statistical modeling of received signals, we show that the information on the number of people is captured in the probability density function (PDF) of the received power measurements, which is then used to estimate the total number of people. We then extend our framework to through-wall scenarios where WiFi transceivers are located outside a building and people are walking inside. We show that the received signal can be modeled as a renewal-type process and show that inter-event times capture the information about the total number of people. We then show several experimental results, using our framework, and show that we can estimate up to $20$ people with a very good accuracy.
Finally, in joint crowd counting and crowd speed estimation, we estimate several occupancy attributes such as the total number of people, their walking speed, and the rate of arrival of people into the area using a pair of standard WiFi links. We further extend our approach to estimate these attributes in the adjacent regions where there may not be any WiFi coverage. We then show several experimental results with various speeds and up to $20$ people and estimate the occupancy attributes. We also show experimental results of our framework in Costco, a retail store.