Student: Ming-Hua Hsieh
Adviser: Kun-chan Lan
The rapid advance of smart phone technology has brought further applications to today's society. Those technologies can provide a huge amount of users' information; therefore many challenging issues are gradually emerging, one of which is indoor positioning. Although there are many indoor positioning technologies developed by utilizing smart phones as the benchmark, those technologies still have a lot of shortages like lacking accuracy and consistency. In recent years, the popularity of iBeacon has gradually be raised, and it has offered a variety of convenient services to many smart phones. In indoor positioning, compared to the Wi-Fi in the traditional system, iBeacon has become a new choice as well. AP, iBeacon has the advantages of low power consumption, small size, and low cost. As far as the low cost is concerned, the number of iBeacons that can be arranged in a specific indoor room can also be increased, so that the signals available under the signal receiving intensity indication system are enhanced to improve the strength of a desired feature. In this study, there is a deep learning approach applied to determine the location of the user. Also, indoor positioning’s received signal strength indicator system has been developed for a long period of time. Currently, the feature extraction methods are not enough to significantly mitigate the effect of the changes on the received signal strength indicator, which lowers the smart phone’s performance of DL-based indoor fingerprinting algorithms. As a result, we have extracted the data of the inertial sensor, calculated the length of the path, judge the direction of the walking parade, and explained the position of the user posterior to each procedure. In addition, it was converted into the signals received by the featured iBeacon. Moreover, since the intensity indicator is integrated, through using iBeacon and inertial sensors in the indoor positioning design, the goal in this paper aims at achieving higher positioning accuracy. As for the experience experiments in real-world environments, they can also show that, deep learning combined with two features in the indoor positioning scheme can be employed to raise the positioning accuracy, while it can raise satisfactory performance in regard of data availability and the amount of cost.