Example application scenario for counting stationary people

Researchers in the laboratory of Yasamin Mostofi, a professor at the University of California, Santa Barbara, used WiFi signals for the first time to calculate a stationary seated crowd without relying on people carrying equipment. The technology can also be counted through walls, and only needs to use wireless transmitters and receivers outside the area of ​​interest where the crowd is located. It has a variety of applications, including smart energy management, crowd control during a pandemic, business planning, and security.  

Mostofi, professor of electrical and computer engineering at UCSB, said: "The method we propose can estimate the number of people seated in an area from the outside." Carry equipment and work through walls."

The proposed method and experimental results appeared at the recently held 19th ACM International Conference on Mobile Systems, Applications and Services (MobiSys). 

In the team's experiment, a WiFi transmitter and a WiFi receiver (both ready-made) are located in an area where many people sit down. The transmitter sends a wireless signal, and its received power is measured by the receiver. By using only such received signal power measurements, the receiver can estimate how many people are present-this estimate closely matches the actual number of people.  

This innovation builds on the previous work of the Mostofi laboratory, which has pioneered the use of daily radio frequency signals (such as WiFi) for sensing since 2009. For example, their 2018 paper showed how WiFi counts the number of mobile people. However, people must move around to be counted.

"Due to the lack of major body movements, counting people who are sitting in a fixed position is a rather challenging problem," Mostofi said. The success of her laboratory in this work is attributed to the new method they developed.

"Although people in a crowd are static, that is, they do not have any major body movements other than breathing, they do not remain still for long periods of time and often perform small on-site natural body movements called fidgeting," Mosto Fei explained. "For example, they may adjust their seat position, cross their legs, check their phones, stretch their bodies or cough, etc."

The researchers proposed that the collective natural restlessness and in-situ movement of the seated crowd carry key information about crowd counting, and for the first time showed how to extract the total irritability and calculate the total number of people based on the crowd. In them. 

“Consider the Crowd Restless Period (CFP), which we define as the duration of at least one person in the WiFi area, and the Crowd Silent Period (CSP), which we define as the period when no one is restless. These periods are easy to receive. The WiFi signal is extracted,” said principal Dr. Belal Korany. Project students. "Intuitively, the more people, the longer the CFP and the shorter the CSP, the more likely it is. Therefore, these periods imply information about the total number of people."

The researchers then developed a new mathematical model that statistically describes the collective fidgeting behavior of stationary people (ie, CFP and CSP), and explicitly correlates them with the number of people seated.

When developing their mathematical formula, they first revealed how this problem is similar to a decades-old queuing theory problem, a seemingly unrelated problem from a completely different field. "Queue theory is a branch of mathematics that studies waiting lines in systems that involve the arrival of customers who need services from entities that include multiple servers," Mostofi said. Then, they showed that CSP is similar to the time that no customer is in a queue with unlimited servers, and CFP is similar to the time that at least one customer is served in such a queue. This allows them to use the mathematical tools of queuing theory to develop a new technique to calculate the total number of stationary personnel.

"We have tested this technology extensively in different locations and tested it with different numbers of people in several different seating configurations," Korany said. The laboratory conducted 47 experiments in four different environments (including through-wall environments) to test their new technology. Up to 10 people were seated and behaving normally, while a pair of WiFi transceivers performed WiFi measurements. Their experiments reflect various occasions, such as attending lectures/demonstrations, watching movies or reading in the library.

Their evaluation results show very high counting accuracy. In a non-wall environment, 96.3% of the time the estimated number of people differs from the real number of people by 0 or 1 person, while in a wall environment it is 90%. Wall setting. Overall, their results show the potential of this new technology for crowd counting in real-world scenarios, such as limiting the total number of people in a crowd during a pandemic.

For a demonstration of this technology, please check out their video at https://www.youtube.com/watch?v=-UTaVmCiUPs

Morie information about the project can be found at https://web.ece.ucsb.edu/~ymostofi/CountingStationaryCrowd

For more information about Mostofi research, please visit https://web.ece.ucsb.edu/~ymostofi/

Sonia Fernandez Sonia(dot)fernandez(at)ucsb(dot)edu

Copyright ©

Regents of the University of California. all rights reserved.