Spectral Human Flow Counting with RSSI in Wireless Sensor Networks

Human flow counting is a fundamental task in public space management. Counting flow correctly may help prevent overcrowding hazards and improve public safety. This study proposes an automated device-free flow counting system by exploiting radio frequency irregularity in a wireless sensor network. As people pass through the line-of-sight between transmitters and receivers, radio frequency transmission is disturbed and received signal strength indicator (RSSI) fluctuates at the receiving ends. Using RSSI fluctuation series, the system infers flow size without patrons' carrying any special devices. A wireless sensor network with HBE-Zigbex motes (IEEE 802.15.4) is set up to conduct experiments. Besides the mean and standard deviation of RSSI fluctuation series, Fourier spectral features are also employed as predictors of a machine learning algorithm. Experimental results show that spectral features improve the prediction accuracy significantly. The proposed method thus provides an alternative solution for the flow counting problem in addition to other video based solutions.