| The popularity of smart phones and the development of Bluetooth technology have made it possible for Passive Entry and Passive Start(PEPS)based on Bluetooth low energy.Traditional PEPS uses smart keys to replace car keys,and Bluetooth Low Energy PEPS is expected to use smartphones to replace smart keys to achieve true keyless entry and start.This paper mainly studies identification technology of the position state of smart terminal in BLE PEPS system.Based on the location fingerprint method and Dempster-Shafer evidence theory,the interior and exterior recognition algorithm are designed.According to the possible actual situation,the experiments of the vehicle application scenario are conducted to verify the recognition performance of the algorithm.The specific research work is as follows:1)Based on the propagation characteristics of RSS,experiments are conducted in the vehicular scenario,clarified the influence of common materials on the vehicle and human body on RSS.Considering the problem that RSS is susceptible to fluctuations caused by environmental factors,a Kalman filter algorithm with good real-time performance is used to suppress noise.Considering about the problem of data redundancy in the fingerprint database in the traditional location fingerprint method,two feature extraction methods,principal component analysis and linear discriminant analysis,are compared.The linear discriminant analysis method is used to reduce the dimensionality of the original RSS data.While ensuring the integrity of valid information,it reduces the dimensions of RSS in offline fingerprint database and test samples.2)The implementation method of location fingerprint technology in vehicle scene is discussed,and the performance evaluation index of vehicle interior and exterior recognition algorithm is established.Based on the RSS distribution of different beacon layout methods,multiple sets of beacon layout schemes are designed,and combined with the application situation,the RSS collection and fingerprint library construction scheme are designed in a targeted manner to obtain relatively complete RSS fingerprint information.On this basis,the location effects of weighted K-nearest neighbor,Bayesian probability,and logistic regression based on location fingerprint positioning are analyzed and compared.3)Considering the limitations of traditional positioning and recognition effects,a differential K-nearest neighbor algorithm is proposed.Through differential processing,the RSS common mode noise is suppressed while enhancing the RSS difference between the inside and outside of the vehicle,thereby achieving a better recognition and positioning effect.The experimental results show that the difference K-nearest neighbor method can identify the position status of terminal in the near-distance position better,while the logistic regression method and the weighted K-nearest neighbor method have a large probability of misjudgment.4)The Dempster-Shafer evidence theory is introduced to fuse the recognition results of different algorithms and improve the robustness of the algorithm.Combining experiments and analysis,different fusion strategies are adopted for different conflict situations to achieve adaptive improvement of the fusion rules of evidence theory.The experimental results show that in the short distance recognition,the fusion algorithm can maintain the recognition accuracy rate of more than 95% in the range of5 cm inside and outside of most windows. |
