In this work, we explore the use of behavioral biometrics when opening doors to enable user recognition on demand. We propose a combination of inertial, capacitance, force, and acoustic sensors embedded in a door for capturing user interaction with the handle. This way, we collect data only when needed, i.e., when the handle is used to open the door. No additional device (e.g., a smartphone) or knowledge is required, enabling a seamless and unobtrusive identification experience for users. We use tangible interaction data captured from 20 participants in two sessions, at least 5 days apart, for building a random forest classifier and an LSTM neural network, and compare and discuss the impact of the sensors on their performance. We found that random forest yields the best accuracy, and performance is better within one session than between sessions. Within one session, a few interactions are sufficient for recognition.      «

In this work, we explore the use of behavioral biometrics when opening doors to enable user recognition on demand. We propose a combination of inertial, capacitance, force, and acoustic sensors embedded in a door for capturing user interaction with the handle. This way, we collect data only when needed, i.e., when the handle is used to open the door. No additional device (e.g., a smartphone) or knowledge is required, enabling a seamless and unobtrusive identification experience for users. We use...     »