Designing datasets for smartphone sensing based localization

Abstract

Since the increasing use of wireless networks, there has been a surge in interest in making use of them for a number of reasons. One of them is localization of mobile devices in both indoor and outdoor settings. Wi-Fi-based localization has more potential than GPS does because of the many factors that reduce the accuracy of GPS positioning. These factors include using GPS in underground environments or in large cities with multi-story buildings, where GPS signals can be blocked or reflected. A lot of data support is needed for Wi-Fi fingerprint-based localisation. The publicly available datasets lack the variety of data collected under different ambient conditions essential to generalize the performance of a localization system. The RSSI of Wi-Fi signals can change depending on the environment. So, The Wi-Fi data was collected and designed under different conditions at two different locations one of which is a ground-level metro station and another one is underground metro. In addition, an inertial sensor dataset based on smartphone inertial sensors (accelerometer, gyroscope, and gravity sensor) was designed to study the movement of smartphones in order to track the users. The details information of metro stations have described, where data was collected. Information on configuration of smartphones have provided which are used for data collection. Some Graphical plot has been shown here in order to assist the process of localization. The designed datasets of JU-MetroLoc have shown a maximum of 78.73% classification accuracy for Random Forest classifier.