People Counting over WiFi
Researchers at the University of California were able to count the number of people sitting still in the classroom using Wi-Fi. The technology does not interact with mobile devices but relies only on mathematical processing of data obtained using standard wireless signal sources.
In the experiments, one transmitter and one receiver of the Wi-Fi signal were used. The devices were installed in an area where there were several people, and by the difference in the received signal strength from each point in space, it was possible to analyze how many people were present in the room.
To count people, a method is used based on the fact that a person cannot be absolutely motionless for a long period of time. A person sitting in place always has small natural body movements, for example, crossing the legs, stretching, coughing, he can also adjust the chair, look at the clock, look into the smartphone, etc. Therefore, despite the fact that counting motionless people is a rather difficult task, it is still possible to record “natural fidgets”.
Below are charts P1, P2, P3, etc., showing micro-movements, in other words, fidgeting of persons - Crowd Fidgeting Periods (CFPs). A non-zero signal indicates "fidget".
Earlier, in 2018, a similar technique made it possible to calculate the number of people moving in a room.
Now a more difficult task has been solved - counting relatively motionless people.
Data processing gives a result close to the actual presence of people in the room. The technology has been tested in different rooms, with different numbers of people and with different seats. A total of 47 experiments were carried out in 4 different environments: lecture, presentation, watching a movie and reading in the library, including detecting people through walls. In general, this approach made it possible to count the number of people in a closed auditorium with an accuracy of 96% and in an auditorium with wall openings with an accuracy of up to 90%. In a small audience, such an error means plus/ minus 1 person.
Key features of the method:
● Does not rely on machine learning, which requires excessive preliminary Wi-Fi data collection, taking into account different numbers of people, seating configurations, and different types of premises;
● based only on a mathematical model;
● does not require people to carry special devices for reading;
● can count people through walls.
Developers see an application for intelligent management of energy consumption, heating, and air conditioning of buildings, for monitoring audience size during a pandemic, as well as for business planning and security purposes.