In large public places such as scenic tourist areas, shopping malls, stations, squares, and so on, there is a wide demand for people counting and pedestrian flow monitoring. Based on the feedback from the pedestrian flow monitoring system, resources can be optimally allocated to maximize social and economic benefits. Moreover, trampling accidents can be avoided because pedestrian guidance is carried out in time. In order to meet these requirements, we propose a method of pedestrian flow monitoring based on the received signal strength (RSS) of wireless sensor networks. This method mainly utilizes the shadow attenuation effect of pedestrians on RF signals of effective links. In this paper, a deployment structure of a radio frequency wireless sensor network is firstly designed to monitor the pedestrians. Secondly, the features are extracted from the wavelet decomposition of RSS signal series with a short time. Lastly, the support vector machine (SVM) algorithm is trained by an experimental data set to distinguish the instantaneous number of pedestrians passing through the monitoring point. The experimental results show that the accuracy is about 92.9% in the context of real-time pedestrian flow monitoring.In large public places such as scenic tourist areas, shopping malls, stations, squares, and so on, there is a wide demand for people counting and pedestrian flow monitoring. Based on the feedback from the pedestrian flow monitoring system, resources can be optimally allocated to maximize social and economic benefits. Moreover, trampling accidents can be avoided because pedestrian guidance is carried out in time. In order to meet these requirements, we propose a method of pedestrian flow monitoring based on the received signal strength (RSS) of wireless sensor networks. This method mainly utilizes the shadow attenuation effect of pedestrians on RF signals of effective links. In this paper, a deployment structure of a radio frequency wireless sensor network is firstly designed to monitor the pedestrians. Secondly, the features are extracted from the wavelet decomposition of RSS signal series with a short time. Lastly, the support vector machine (SVM) algorithm is trained by an experimental data set to distinguish the instantaneous number of pedestrians passing through the monitoring point. The experimental results show that the accuracy is about 92.9% in the context of real-time pedestrian flow monitoring.