CLSLoc Deep Learning Indoor Positioning Model Based on Wi-Fi RSSI Fingerprint

Abstract

Indoor positioning technologies provide a wide range of application demands on large-scale buildings with the continuous development of modern society. Among them, the most popular solution is based on fingerprint recognition, which mainly uses signals from Wi-Fi access points and Bluetooth Low Energy (BLE) beacons. In the implementation of indoor positioning systems based on fingerprint recognition, it is divided into three stages: environmental description, model establishment, and evaluation. The focus of this dissertation lies in the model establishment and evaluation phase of indoor positioning systems (IPS) based on fingerprints. Firstly, we propose an indoor positioning model, namely CLSLoc (CNN-LSTM-Stacking Localization), which significantly improves the localization accuracy. Secondly, we attempt to reduce the inference time of the model by employing data dimensionality reduction techniques and compare the Principal Component Analysis (PCA) and Kernel Principal Component Analysis (KPCA) dimensionality reduction methods. We tested CSLoc's ability to locate buildings and floors and compared it with several models. We also tested the data sets after PCA and KPCA dimensionality reduction and the data sets without dimensionality reduction to compare the inference speed of the models. We evaluated CLSLoc and several other methods on the UJIIndoorLoc and TUT series datasets. The experimental results show that CLSLoc has the highest positioning accuracy (99.57%) compared to other existing methods. At the same time, it has been verified that PCA and KPCA methods can effectively reduce the inference time (around 50%) of the model.

??????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????? ????????????????????????????????????????????????????????????? ????????????????????????????????????????????-???????????????? ?????????????????????????????????????????????????????????? ???????????????????????: ?????????????????????? ???????????????? ??????????????? ???????????????????????????????????????????????????????????????????????????????????? (Indoor Positioning Systems; IPS) ???????????????????? ????????? ???????????????????????????????????? CLSLoc (CNN-LSTM-Stacking Localization) ????????????????????????????????????????????????? ???????????? ?????????????????????????????????????????????????????????????????????????????????????????????????????????????????? (Principal Component Analysis; PCA) ????????????????????????????????????? (Kernel Principal Component Analysis; KPCA) ??????????????????????? CLSLoc ?????????????????????????? ?????????????????????????????? ???????????????????????????????????????????? PCA ??? KPCA ??????????????????????????????????????????????????????????????????????????? ???????????? CLSLoc ????????????????? ??????????????????? UJIIndoorLoc ??? TUT ???????????????????????? CLSLoc ?????????????????????????????????? (99.57%) ????????????????????????? ????????? ????????????????????????????? PCA ??? KPCA ??????????????????????? (?????? 50%) ????????????????????????????????