A device-free Wi-Fi sensing method for pedestrian monitoring using channel state information

Pedestrian detection accuracy strongly impacts the effectiveness and reliability of intelligent pedestrian-related control systems. Traditional sensing technologies usually sense pedestrians based on the reflected signal of the transmitted infrared ray, sound wave, and electromagnetic wave which only can count the number of times that pedestrians passing a line of sight (LoS) but the moving feature monitoring, e.g., moving direction, speed, etc. For pedestrian monitoring based on computer vision-based sensing technology, the level of errors is relatively large and highly sensitive to environmental factors, such as illumination, weather conditions, and occlusion. Wi-Fi channel state information (CSI) represents the amplitudes and phases information for orthogonal frequency-division multiplexing (OFDM) subcarriers, which is mainly impacted by the static environment and moving object in surrounding areas. Previously, scholars utilized Wi-Fi CSI to analyzed multiple microscopic human movements, e.g., gesture, gait, and fall action in the indoor environment, but no application in the outdoor environment for pedestrian monitoring. The main objective of this research is to demonstrate the feasibility and reliability of the Wi-Fi CSI-based sensing method for pedestrian existence and moving direction recognition. The impacts of the CSI signal sampling ratio on the detection accuracy was investigated as well. The experiments were conducted in both indoor and outdoor environments. According to the results, the accuracy of pedestrian existence detection based on the data of the 100 Hz sampling ratio achieved 99.23% accuracy and 0.26% fast positive rate. For the moving direction recognition, the detection accuracy in the indoor environment achieved 100% and 96.92% for two directions, and got 92.21% and 93.51% in the outdoor environment. The findings of this research demonstrate the proposed Wi-Fi CSI signal is highly effective for pedestrian existence detection and moving direction recognition. The future research will continue in pedestrian moving speed estimation, overlapped pedestrian identification, and pedestrian, bicyclists, and wheelchair classification.