Exploiting Radio Irregularity in Wireless Networks for Automated People Counting
Wireless devices exist almost everywhere in our daily life. Wireless communications, which is an integral part of wireless devices, suffers from radio irregularity – a phenomenon referring to radio waves being selectively absorbed, reflected or scattered by objects in their paths, e.g., human bodies that comprises liquid, bone and flesh. Radio irregularity is often treated as a major challenge for wireless communication. However, we aim to take advantage of the phenomenon of radio irregularity to provide a cost-effective approach for automated people counting. People counting is extensively used for intelligence-gathering to be used in forecasting, resource allocation and safety-related applications like crowd control. Existing people counting techniques use light, infrared, or thermal energy for human movement detection. However there have major limitations, for example the visible light camera and infrared sensors do not penetrate smoke or obstacles such as wall and furniture. Also, a large deployment of these devices is costly owing to the use of specialized sensors.
We propose an automated people counting system using the radio irregularity phenomenon of existing wireless infrastructure with minimal additional hardware and installation costs. This thesis presents an experimental study to demonstrate how radio signal fluctuations arising from radio irregularity can be used to provide a simple low-cost alternative to dedicated sensing systems for indoor automated people counting. Firstly, we study the effect on received signal strength with human motion interference on radio signals. Then we propose and evaluate the performance of three approaches, namely, overcomplete dictionary based pattern recognition (OCPR) approach, probability density approach and standard deviation approach. With high accuracy of motion detection, we then focus on the design of automated people counting system using the proposed detection approach. To differentiate the number of people, we apply discriminant analysis which is a statistical method to perform classification based on independent variables. We validated the proposed people counting system by conducting experiments under both controlled and uncontrolled environments and show that we are able to achieve high accuracy in counting up to five people in groups with no specific formation.