Demonstrations and people-counting based on Wifi probe requests

At demonstrations, counting people based on the Wifi signals that their phones emit, may be a flexible alternative when conventional methods fail. However, at public events like that, such Wifi signals do not only arrive from the targeted crowd but also from bystanders that happen to be in the same area. This raises the question of suitable filtering mechanisms and their impact on the counting setup as well as the accuracy of the method as such. This paper explores distance- and time-based filters and proposes filtering by recognition to estimate the size of a marching crowd. It deploys the resulting setup at two demonstrations each with over a thousand people. The experiments show that the signaling behaviour of different phones varies significantly and that the count of Wifi probe requests after filtering represents a small fraction of the actual attendance.