AITouch - Data

Touch-based interaction data were captured for 37 volunteers through an ad-hoc mobile application developed to this purpose. Different scenes were proposed so that the most common gestures were captured in order to study behavioral patterns, mostly oriented to authentication purposes. Every user was appropriately informed about the information gathered and its intended use, requesting their consent. During the data acquisition process itself, all the users used the same device, a Samsung Tab S4, to guarantee that the data was not dependent on the device used. Moreover, to ensure that the user sticks to the guidelines provided and to guarantee the quality of the data being gathered, we monitored every user during such a process.
On the one hand, raw data is available in JSON format, which includes essentially positions, timestamps and contact surface measurements for each active finger. This results in around 24,000 events (700,000 individual measurements) as they were recorded by the application, with no further processing.
On the other hand, we also provide some features derived from motion and geometric characteristics, which have already been sanitized (spurious or outlier samples were removed) and preprocessed (average values are provided for each event, rather than for individual measurements). In addition, the code used to extract and sanitize such features is also available so that the user can reproduce, adapt or even extend them to conduct their own research.