Database Reference
In-Depth Information
responding mobile devices within the scanners communication range. Every
time a device is detected, its MAC address, COD (Class of Device) code,
and the timestamp of the detection are registered. Additionally, the received
signal strength intensity (RSSI) of the inquiry response is logged. This inten-
sity value is inferred from the received power level with which the response
packet was detected by the scanner and is theoretically negatively correlated
with the distance between the scanner and the detected device. Because some
users include personal information in the friendly name of the detected device
(name, phone number, etc.), it is not registered to safeguard privacy. The inquiry
phase does not require an active connection between the scanner and the mobile
device, so the methodology does not necessitate any cooperation of the tracked
individual.
By placing Bluetooth scanners at different strategic locations, meaningful
trajectories generated by mobile devices (and correspondingly by their owners)
can be reconstructed. Because of the complex environmental setting and the
resulting unpredictability of the propagation of Bluetooth signals, positioning is
currently done through the
proximity
principle, where the position of a detected
mobile device is approximated to the point position of the scanner by which it is
detected. The strategic locations of the scanners are used to semantically enrich
the resulting trajectories, which then become geo-localized semantic trajectories.
As with any other form of sparsely sampled (sometimes also called
episodic
)
movement data, the locations of mobile devices that are not within range of any
scanner are unknown.
The spatial granularity of the resulting trajectories ultimately depends on the
detection range of the Bluetooth scanners, and on the number and coverage of
Bluetooth scanners within the study area. In theory, the detection range depends
on the power class of the Bluetooth device (Class 1: 100 m, Class 2: 10 m,
Class 3: 1 m). In practice, however, this range is variable due to environmental
factors influencing (blocking, reflecting, etc.) radio signals leading to a detection
region with a fuzzy border. The temporal granularity cannot be predicted either
because the Bluetooth scanners register detections whenever they arrive instead
of using a fixed sampling interval. Devices within a direct line of sight with a
sensor will usually lead to new detections every few seconds.
14.2.2 Preprocessing and Software
The raw tracking data consist of log files - named after the combination of
the scanner and the MAC address of the sensor - containing log lines with
the following format:
timestamp of detection, MAC address of the detected
device, COD code of the detected device, RSSI of detection
. In order to obtain
a compressed data set, the scanners are programmed to create a second set
