| With the rapid development of communication,computing and control technologies,wireless sensor networks have been more and more widely used.Indoor positioning technology has a broad application prospect.Wi Fi(Wireless Fidelity)signals are free of additional equipment,low-cost,easy to implement and other characteristics,which have been widely studied and practiced by people.Its application scope includes indoor places such as shopping malls,hospitals,airports,museums,etc.,which can provide people with precise indoor positioning and navigation services.In indoor environments,there is a problem of signal strength variation when receiving Wi Fi signals.This is mainly due to the instability of signal strength values caused by interference such as reflection and scattering.To address this issue,researchers need to preprocess the raw data and construct an accurate location fingerprint library to improve the accuracy of indoor Wi Fi positioning.By studying preprocessing and fingerprint library construction algorithms,the feasibility and application range of Wi Fi indoor positioning technology can be further improved.The specific research content is as follows:Optimization model for RSS(Received Signal Strength)fingerprint collection based on normality detection.This article proposes a normality detection method based on kurtosis and skewness testing to test the skewness and kurtosis of RSS samples collected in the offline stage.If the collected RSS samples meet the normal model,the normal distribution method is used to estimate the probability density of fingerprint points.If the normal model is not satisfied,the Gaussian kernel density function is used to estimate the probability density and filter the high probability signal.After smoothing the RSS sample data,calculate the average value and store it in the fingerprint database.The experimental results show that the method proposed in this article can minimize the error caused by RSSI fluctuations,thereby improving the positioning accuracy in the positioning stage.A WiFi target localization model based on LWKNN.This article proposes a target localization method based on LWKNN in the online localization stage.This method first performs preliminary position estimation using WKNN(Weighted K-nearest Neighbor Algorithm).Next,the historical positioning information is used as feature information to establish an LSTM(Long Short Term Memory networks)model.Through training,the predicted results are fed back to the subsequent target position positioning,which is used to exclude position points with significant deviation from the predicted trajectory.Finally,the target positioning position is output.This method combines trajectory prediction models with fingerprint localization for the first time,which can effectively utilize the historical information of target trajectories.The experimental results show that the indoor Wi Fi fingerprint localization method based on LWKNN proposed in this paper effectively reduces errors and improves the accuracy of target localization. |
