Research On Human Flow Monitoring Method Based On WiFi Channel Characteristics

In recent years,with the prosperity and development of economy and culture,people often gather in public places such as tourist attractions and subway stations.Crowd management has become more and more important.Human flow monitoring is the basis for measures such as human flow control and rescue and is used to count the number of people in a specific area.This paper proposes a human flow monitoring method based on commercial Wi Fi.The method has the advantages of easy installation,low cost and strong applicability comparing with the traditional human flow monitoring methods.There is no need for the monitored personnel to wear any intelligent device at the same time.It achieves the goal of passive monitoring truly and the purpose of intelligently perception human flow.Under the wave of big data and artificial intelligence,wireless perception has become a hot research topic of domestic and foreign experts.Wireless perception technology is an emerging technology that infers the state and actions of a target person by analyzing the impact of limbs on surrounding wireless signals.Based on this technology,this paper proposes a practical human flow monitoring scheme,called Wi-HFM,a human flow monitoring system based on commercial Wi Fi and deep learning networks.Specifically,in the data collection stage,using commercial Wi Fi to build an AP mode data collection platform with one transmitting antenna and three receiving antennas.We organize personnel to carry out human flow simulation experiments and collecting multiple sets of data.In the data processing and analysis stage,the fine-grained information in the channel state information measurement value is fully utilized to preprocess the amplitude information and phase information respectively,including the selection of antenna pairs and data packets,and data cleaning.In order to better characterize the influence of the number of people on the wireless signal,it is necessary to obtain the optimal characteristic signal.After using PCA to reduce the dimension of the data extracted from the time domain and wavelet domain features,the effect of the signal in the time domain and time-frequency domain is analyzed through the characteristics.Finally,the phase difference in the time domain is determined as the identification characteristic signal.In the stage of constructing the human flow monitoring model,deep learning can be used to learn the characteristics of the latent features of the data,and the convolutional neural network that can mine the latent features and the recurrent neural network that can extract the time domain features of the signal are used to construct the CLDNN model to monitor the human flow.In this paper,a large number of experimental data samples are collected by building indoor and outdoor simulated environments.In order to evaluate the effect of the human flow monitoring system,the influence of various parameters in the model and experimental scenarios on the human flow monitoring was studied,and the optimal parameters were selected for the human flow monitoring.The experimental results show that the method in this paper is an practical human flow monitoring method that can protect personal privacy,which can provide a certain research reference for the development of the field of people statistics.