| With the rapid development of mobile Internet,the demand for location-based services is increasing.Indoor localization technology holds out a great prospect for further expansion of LBS in such areas as hospital telenursing,asset management and industrial production,therefore,indoor localization technology has become the research hotspot in recent years.Among the technologies used in indoor localization,bluetooth low energy technology has become a research hotspot because it has the characteristics of low cost,low power consumption and high localization accuracy.But the existing localization solutions are difficult to achieve a good balance between localization accuracy and cost.In this thesis,the localization accuracy and system cost are considered comprehensively,and a high-precision indoor localization method based on long-short-term memory network is proposed.The localization scheme based on BLE Received Signal Strength Indication(RSSI)is determined,by improving the existing fingerprint localization algorithm,the indoor localization system is designed and implemented.The hardware of the system is low-cost,and it has high localization accuracy.The main research work in the thesis includes the following aspects:First,the principle of the system is analyzed,and the overall system design is divided into four parts: design of RSSI data enhancement algorithm,design of RSSI filter algorithm,design of the indoor localization algorithm and the implementation and test of the overall system.Then the software part of the system is designed,which is divided into two parts: the mobile terminal and server side.The main idea is to complete the establishment of the fingerprint database and the calculation of the localization coordinates on the server side with strong computing power,and the mobile terminal is mainly responsible for the collection of RSSI data and the display of the results.The server side establishes a connection with the mobile terminal through the wireless network based on the TCP/IP protocol.Then,the algorithms of the system are designed.First,design the RSSI data enhancement algorithm.In the offline sampling stage of the fingerprint localization algorithm,using generative adversarial networks to generate RSSI data with the real RSSI data collected,the generated RSSI data has the similar distribution compared with the real data.By using the generated data to replace part of the real data to build a fingerprint database,the purpose of reducing the workload of fingerprint data collection is achieved,and it is convenient to the rapid establishment of an offline fingerprint database.Then design the RSSI filter algorithm,the Kalman filter algorithm is studied to eliminate the interference of noise in the environment to RSSI data and improve the localization accuracy of subsequent indoor localization algorithm.Then,design the indoor localization algorithm,considering that the recurrent neural networks are suitable for processing time series data,the output can be calculated non-linearly based on the input.An indoor localization algorithm based on Long ShortTerm Memory(LSTM)is proposed,which utilizes the temporal information in the localization process to improve the localization accuracy.The neural network in the proposed algorithm consists of fully-connected and LSTM layers,and the input is a time-series of RSSI and output is the estimated location at the latest time in the input.After the network model is trained on the fingerprint database constructed by the RSSI data collected in the actual environment,the average prediction accuracy of the algorithm is significantly improved compared with that of the multi-layer perceptron and other localization algorithms.Finally,complete the construction of the relevant environment and the overall realization of the system.The localization accuracy of the system is tested at different locations in the indoor environment.The results show that the average localization accuracy of the system reaches 1.31 m,which proves the practical value of the system. |
