Research On Indoor WIFI Localization Algorithm For Sparse Samples

With the popularization of smart terminals and the development of mobile Internet,people's demand for location-based services in indoor environments has surged.As WiFi is widely deployed in indoor environments,this provides a lot of convenience for WiFi-based indoor localization.The WiFi-based fingerprinting localization method has the characteristics of strong universality and wide coverage,and has been widely studied and adopted,but there are problems such as a large amount of data collection work and a lot of time-consuming.Aiming at the above problems,this paper proposes an indoor localization algorithm for sparse samples.Firstly,this paper designs a sparse sample expansion algorithm PGSE based on Principal Component Analysis(PCA)and Gaussian Process Regression(GPR).Based on the WiFi signal of some sample points in the localization area,the WiFi signal of other sample points is generated through the expansion algorithm,and then the data is expanded according to the generated WiFi signal and the fingerprint database of the area is constructed.Secondly,for two different types of data without time series information and with time series information in a real environment,the localization algorithm EDL based on Deep Neural Networks and the localization algorithm DLTSL based on Long ShortTerm Memory networks are designed respectively.Finally,this paper carried out a lot of experiments in real scenes to verify the effectiveness of the expansion algorithm and the localization algorithm.The experimental results show that the data expanded by PGSE is compared with the original data,the corresponding similarity and the localization accuracy are very closely;for the data without timing information,the average localization error of the EDL localization algorithm is 2.7 meters,and the probability of localization error within 3 meters is 80%;for the data with timing information,the average localization error of the DLTSL localization algorithm is 3.2 meters,and the probability of localization error within 2.5 meters is 80%.In summary,the sparse sample-oriented expansion algorithm proposed in this paper reduces the workload of collecting data while ensuring the localization accuracy.Compared with other localization algorithms,the localization algorithm proposed in this paper not only improves the accuracy of localization,but also is suitable for a variety of different localization scenarios.