Research Of Crowd Estimation Method Based On WiFi Signal Measurement

With the development of society and progress of science and technology,the number of people in a specific region is a very useful information in many real-life applications,crowd estimation method attract a lot of scholars' research interest.The former human-vision based method is inefficient and often results in huge error,so many new crowd estimation methods have been proposed in recent years,among these methods,WiFi based method is most promising,it does not strict to LOS,not affected by lighting conditions,does not require users to attach any device,and can be applied on the existing WiFi infrastructures.It's low-cost and easy to deploy.But existing WiFi based method still has some shortcomings,such as its accuracy is not very high and it can only estimate the moving people.So this paper focus on WiFi based crowd estimation methods,trying to overcome the existing methods' shortcomings.The main contribution of this paper is as follows:1.A breath-detecting based static target detecting method is proposed.Human respiration is accompanied with expansion and shrink of the chest,this kind of faint movement in chest will result in the fluctuation in CSI,through phase sanitization,outlier removing,low-pass filtering and Fourier transform,we can extract and detect this kind of fluctuation in CSIs phase and amplitude to detect breath activity,then we can detect the static target.The innovations of this method are: 1)We apply breath detecting in crowd estimation to detect static people.2)The phase information of CSI is used in breath detecting,thus increasing the amount of available information greatly and improving the accuracy and robustness of breath detecting.Experiment results show that the proposed method can detect the static target with extremely high accuracy and the error between breath frequency got by the proposed method and real breath frequency is less than 3%,2.A CSI based moving people counting method is proposed.Through the experiment we find that under different number of moving people conditions,there is a significant difference in the severity of CSI volatility,it roughly increases along with the increase of the number of moving people.In order to describe the CSI volatility under different number of moving people,we adopt sliding window mechanism to calculate the variance of CSI,the variance is used as an attribute of CSI.In training stage,SVM is used to learn the difference of the CSI's variance under different number of moving people and get trained SVM classifier,in the following monitoring stage,the trained SVM classifier is used to perform moving people counting.Experiment results show that more than 90% results are accurate and the maximum estimation error is less than 1 people.